chore: update nix node_modules hashes

fix(opencode): select Kimi prompt by provider (#42161 )
Co-authored-by: Aiden Cline <rekram1-node@users.noreply.github.com>
2026-08-12 20:50:01 -04:00 · 2026-08-12 22:36:40 +00:00 · 2026-08-12 17:29:03 -05:00 · 2026-08-12 17:20:23 -05:00 · 2026-08-12 17:20:07 -05:00 · 2026-08-12 17:19:17 -05:00
87 changed files with 1067 additions and 319 deletions
@@ -10,6 +10,7 @@ kitlangton
 kommander
 ludvigrask
 MrMushrooooom
+neriousy
 nexxeln
 R44VC0RP
 rekram1-node
@@ -35,6 +35,7 @@ jobs:

      - run: bun sst deploy --stage=${{ github.ref_name }}
        env:
+          GITHUB_TOKEN: ${{ github.token }}
          CLOUDFLARE_API_TOKEN: ${{ secrets.CLOUDFLARE_API_TOKEN }}
          PLANETSCALE_SERVICE_TOKEN_NAME: ${{ secrets.PLANETSCALE_SERVICE_TOKEN_NAME }}
          PLANETSCALE_SERVICE_TOKEN: ${{ secrets.PLANETSCALE_SERVICE_TOKEN }}
@@ -4,7 +4,7 @@ import { tool } from "@opencode-ai/plugin"
 const TEAM = {
  tui: ["kommander", "simonklee"],
  desktop_web: ["Hona", "Brendonovich"],
-  core: ["jlongster", "rekram1-node", "nexxeln", "kitlangton"],
+  core: ["jlongster", "rekram1-node", "neriousy", "nexxeln", "kitlangton"],
  inference: ["fwang", "MrMushrooooom", "starptech"],
  windows: ["Hona"],
 } as const
@@ -29,7 +29,7 @@
    },
    "packages/app": {
      "name": "@opencode-ai/app",
-      "version": "1.18.16",
+      "version": "1.18.17",
      "dependencies": {
        "@corvu/drawer": "catalog:",
        "@dnd-kit/abstract": "0.5.0",
@@ -96,7 +96,7 @@
    },
    "packages/cli": {
      "name": "@opencode-ai/cli",
-      "version": "1.18.16",
+      "version": "1.18.17",
      "bin": {
        "lildax": "./bin/lildax.cjs",
      },
@@ -144,7 +144,7 @@
    },
    "packages/codemode": {
      "name": "@opencode-ai/codemode",
-      "version": "1.18.16",
+      "version": "1.18.17",
      "dependencies": {
        "acorn": "8.15.0",
        "effect": "catalog:",
@@ -158,7 +158,7 @@
    },
    "packages/console/app": {
      "name": "@opencode-ai/console-app",
-      "version": "1.18.16",
+      "version": "1.18.17",
      "dependencies": {
        "@cloudflare/vite-plugin": "1.15.2",
        "@ibm/plex": "6.4.1",
@@ -194,7 +194,7 @@
    },
    "packages/console/core": {
      "name": "@opencode-ai/console-core",
-      "version": "1.18.16",
+      "version": "1.18.17",
      "dependencies": {
        "@aws-sdk/client-sts": "3.782.0",
        "@jsx-email/render": "1.1.1",
@@ -221,7 +221,7 @@
    },
    "packages/console/function": {
      "name": "@opencode-ai/console-function",
-      "version": "1.18.16",
+      "version": "1.18.17",
      "dependencies": {
        "@ai-sdk/anthropic": "3.0.82",
        "@ai-sdk/openai": "3.0.48",
@@ -243,7 +243,7 @@
    },
    "packages/console/mail": {
      "name": "@opencode-ai/console-mail",
-      "version": "1.18.16",
+      "version": "1.18.17",
      "dependencies": {
        "@jsx-email/all": "2.2.3",
        "@jsx-email/cli": "1.4.3",
@@ -267,7 +267,7 @@
    },
    "packages/console/support": {
      "name": "@opencode-ai/console-support",
-      "version": "1.18.16",
+      "version": "1.18.17",
      "dependencies": {
        "@cloudflare/vite-plugin": "1.15.2",
        "@opencode-ai/console-core": "workspace:*",
@@ -287,7 +287,7 @@
    },
    "packages/core": {
      "name": "@opencode-ai/core",
-      "version": "1.18.16",
+      "version": "1.18.17",
      "bin": {
        "opencode": "./bin/opencode",
      },
@@ -381,7 +381,7 @@
    },
    "packages/desktop": {
      "name": "@opencode-ai/desktop",
-      "version": "1.18.16",
+      "version": "1.18.17",
      "dependencies": {
        "@zip.js/zip.js": "2.7.62",
        "drizzle-orm": "catalog:",
@@ -435,7 +435,7 @@
    },
    "packages/effect-drizzle-sqlite": {
      "name": "@opencode-ai/effect-drizzle-sqlite",
-      "version": "1.18.16",
+      "version": "1.18.17",
      "dependencies": {
        "drizzle-orm": "catalog:",
        "effect": "catalog:",
@@ -449,7 +449,7 @@
    },
    "packages/effect-sqlite-node": {
      "name": "@opencode-ai/effect-sqlite-node",
-      "version": "1.18.16",
+      "version": "1.18.17",
      "dependencies": {
        "effect": "catalog:",
      },
@@ -461,7 +461,7 @@
    },
    "packages/enterprise": {
      "name": "@opencode-ai/enterprise",
-      "version": "1.18.16",
+      "version": "1.18.17",
      "dependencies": {
        "@hono/standard-validator": "catalog:",
        "@opencode-ai/core": "workspace:*",
@@ -493,7 +493,7 @@
    },
    "packages/function": {
      "name": "@opencode-ai/function",
-      "version": "1.18.16",
+      "version": "1.18.17",
      "dependencies": {
        "@octokit/auth-app": "8.0.1",
        "@octokit/rest": "catalog:",
@@ -509,7 +509,7 @@
    },
    "packages/http-recorder": {
      "name": "@opencode-ai/http-recorder",
-      "version": "1.18.16",
+      "version": "1.18.17",
      "dependencies": {
        "@effect/platform-node": "4.0.0-beta.83",
        "@effect/platform-node-shared": "4.0.0-beta.83",
@@ -540,7 +540,7 @@
    },
    "packages/llm": {
      "name": "@opencode-ai/llm",
-      "version": "1.18.16",
+      "version": "1.18.17",
      "dependencies": {
        "@opencode-ai/schema": "workspace:*",
        "@smithy/eventstream-codec": "4.2.14",
@@ -559,7 +559,7 @@
    },
    "packages/opencode": {
      "name": "opencode",
-      "version": "1.18.16",
+      "version": "1.18.17",
      "bin": {
        "opencode": "./bin/opencode",
      },
@@ -690,7 +690,7 @@
    },
    "packages/plugin": {
      "name": "@opencode-ai/plugin",
-      "version": "1.18.16",
+      "version": "1.18.17",
      "dependencies": {
        "@ai-sdk/provider": "3.0.8",
        "@opencode-ai/sdk": "workspace:*",
@@ -766,7 +766,7 @@
    },
    "packages/sdk/js": {
      "name": "@opencode-ai/sdk",
-      "version": "1.18.16",
+      "version": "1.18.17",
      "dependencies": {
        "cross-spawn": "catalog:",
      },
@@ -781,7 +781,7 @@
    },
    "packages/server": {
      "name": "@opencode-ai/server",
-      "version": "1.18.16",
+      "version": "1.18.17",
      "dependencies": {
        "@opencode-ai/core": "workspace:*",
        "@opencode-ai/protocol": "workspace:*",
@@ -796,7 +796,7 @@
    },
    "packages/session-ui": {
      "name": "@opencode-ai/session-ui",
-      "version": "1.18.16",
+      "version": "1.18.17",
      "dependencies": {
        "@kobalte/core": "catalog:",
        "@opencode-ai/client": "file:../app/vendor/opencode-ai-client-1.17.13-v2.tgz",
@@ -836,7 +836,7 @@
    },
    "packages/slack": {
      "name": "@opencode-ai/slack",
-      "version": "1.18.16",
+      "version": "1.18.17",
      "dependencies": {
        "@opencode-ai/sdk": "workspace:*",
        "@slack/bolt": "^3.17.1",
@@ -849,7 +849,7 @@
    },
    "packages/stats/app": {
      "name": "@opencode-ai/stats-app",
-      "version": "1.18.16",
+      "version": "1.18.17",
      "dependencies": {
        "@ibm/plex": "6.4.1",
        "@kobalte/core": "catalog:",
@@ -883,7 +883,7 @@
    },
    "packages/stats/core": {
      "name": "@opencode-ai/stats-core",
-      "version": "1.18.16",
+      "version": "1.18.17",
      "dependencies": {
        "@aws-sdk/client-athena": "3.933.0",
        "@planetscale/database": "1.19.0",
@@ -902,7 +902,7 @@
    },
    "packages/stats/server": {
      "name": "@opencode-ai/stats-server",
-      "version": "1.18.16",
+      "version": "1.18.17",
      "dependencies": {
        "@aws-sdk/client-firehose": "3.933.0",
        "@effect/platform-node": "catalog:",
@@ -944,7 +944,7 @@
    },
    "packages/tui": {
      "name": "@opencode-ai/tui",
-      "version": "1.18.16",
+      "version": "1.18.17",
      "dependencies": {
        "@opencode-ai/core": "workspace:*",
        "@opencode-ai/plugin": "workspace:*",
@@ -971,7 +971,7 @@
    },
    "packages/ui": {
      "name": "@opencode-ai/ui",
-      "version": "1.18.16",
+      "version": "1.18.17",
      "dependencies": {
        "@kobalte/core": "catalog:",
        "@pierre/diffs": "catalog:",
@@ -1022,7 +1022,7 @@
    },
    "packages/web": {
      "name": "@opencode-ai/web",
-      "version": "1.18.16",
+      "version": "1.18.17",
      "dependencies": {
        "@astrojs/cloudflare": "12.6.3",
        "@astrojs/markdown-remark": "6.3.1",
@@ -1075,6 +1075,7 @@
    "@silvia-odwyer/photon-node@0.3.4": "patches/@silvia-odwyer%2Fphoton-node@0.3.4.patch",
    "@npmcli/agent@4.0.2": "patches/@npmcli%2Fagent@4.0.2.patch",
    "solid-js@1.9.10": "patches/solid-js@1.9.10.patch",
+    "@ai-sdk/groq@3.0.31": "patches/@ai-sdk%2Fgroq@3.0.31.patch",
    "@ai-sdk/google@3.0.73": "patches/@ai-sdk%2Fgoogle@3.0.73.patch",
    "pacote@21.5.0": "patches/pacote@21.5.0.patch",
    "@dnd-kit/dom@0.5.0": "patches/@dnd-kit%2Fdom@0.5.0.patch",
@@ -181,6 +181,16 @@ const statsSyncConfig = new sst.Linkable("StatsSyncConfig", {
  },
 })

+const r2SqlAuthToken = new sst.Secret("R2SqlAuthToken")
+const r2Sql = new sst.Linkable("R2Sql", {
+  properties: {
+    accountId: "15d29c8639fd3733b1b5486a2acfd968",
+    bucket: `platform-${$app.stage}-lake`,
+    namespace: "inference",
+    table: "generation",
+  },
+})
+
 export const statSync = new sst.aws.Service("StatsSyncService", {
  cluster: lakeCluster,
  architecture: "arm64",
@@ -193,7 +203,9 @@ export const statSync = new sst.aws.Service("StatsSyncService", {
    dockerfile: "packages/stats/server/Dockerfile",
  },
  command: ["bun", "src/stat-sync.ts"],
-  link: [database, inferenceEvent, statsSyncConfig],
+  // Keep the legacy Athena link and IAM permissions during the first R2-backed
+  // release so reverting the application code remains a one-deploy rollback.
+  link: [database, inferenceEvent, r2Sql, r2SqlAuthToken, statsSyncConfig],
  permissions: lakeQueryPermissions,
  scaling: {
    min: 1,
@@ -1,8 +1,8 @@
 {
  "nodeModules": {
-    "x86_64-linux": "sha256-uduwrM143NDSc+tXsi4lVVfoMll2a3BDHRUjuO7GB68=",
-    "aarch64-linux": "sha256-6DUda78XdXY6DP86lIUkweSjys3iG4Y4mo1PiaNuXbg=",
-    "aarch64-darwin": "sha256-AkJwfLULLZVwwz+XU1QcFUZoIS7oVPCn+n/MXEaxrqE=",
-    "x86_64-darwin": "sha256-hAxKGdiITTxQ2uujQt6prNjo3NxGAMMeo+9HlMWK6GU="
+    "x86_64-linux": "sha256-TNwKfqxD83UpZuCKN8FdEWN+CcQUP9CkCQSLGNqR/sA=",
+    "aarch64-linux": "sha256-qzvOJZzmq2QhlauElw8GwgQnCPHdhexI52L0md5zrxQ=",
+    "aarch64-darwin": "sha256-ZzoyLayOFfcYUAg35ZbZ2WapxDdd9IUWqy2xkxZH4QM=",
+    "x86_64-darwin": "sha256-maP/qLeaC3q8VcmNIPyIKlnplxFXJ7ULho3v21/16Mw="
  }
 }
@@ -159,6 +159,7 @@
    "@modelcontextprotocol/sdk@1.29.0": "patches/@modelcontextprotocol%2Fsdk@1.29.0.patch",
    "effect@4.0.0-beta.83": "patches/effect@4.0.0-beta.83.patch",
    "@tanstack/virtual-core@3.17.3": "patches/@tanstack%2Fvirtual-core@3.17.3.patch",
-    "@ai-sdk/openai-compatible@2.0.41": "patches/@ai-sdk%2Fopenai-compatible@2.0.41.patch"
+    "@ai-sdk/openai-compatible@2.0.41": "patches/@ai-sdk%2Fopenai-compatible@2.0.41.patch",
+    "@ai-sdk/groq@3.0.31": "patches/@ai-sdk%2Fgroq@3.0.31.patch"
  }
 }
@@ -1,6 +1,6 @@
 {
  "name": "@opencode-ai/app",
-  "version": "1.18.16",
+  "version": "1.18.17",
  "description": "",
  "type": "module",
  "exports": {
@@ -1,7 +1,7 @@
 {
  "$schema": "https://json.schemastore.org/package.json",
  "name": "@opencode-ai/cli",
-  "version": "1.18.16",
+  "version": "1.18.17",
  "type": "module",
  "license": "MIT",
  "bin": {
@@ -1,7 +1,7 @@
 {
  "$schema": "https://json.schemastore.org/package.json",
  "name": "@opencode-ai/codemode",
-  "version": "1.18.16",
+  "version": "1.18.17",
  "description": "Effect-native confined code execution over schema-described tools",
  "private": true,
  "type": "module",
@@ -1,6 +1,6 @@
 {
  "name": "@opencode-ai/console-app",
-  "version": "1.18.16",
+  "version": "1.18.17",
  "type": "module",
  "license": "MIT",
  "scripts": {
@@ -139,7 +139,7 @@ export async function handler(
    if (
      authInfo &&
      opts.modelList === "lite" &&
-      modelInfo.id === "deepseek-v4-flash" &&
+      ["deepseek-v4-flash", "deepseek-v4-pro"].includes(modelInfo.id) &&
      !allowedRegions?.includes("cn")
    )
      throw new RegionError(
@@ -1,7 +1,7 @@
 {
  "$schema": "https://json.schemastore.org/package.json",
  "name": "@opencode-ai/console-core",
-  "version": "1.18.16",
+  "version": "1.18.17",
  "private": true,
  "type": "module",
  "license": "MIT",
@@ -1,6 +1,6 @@
 {
  "name": "@opencode-ai/console-function",
-  "version": "1.18.16",
+  "version": "1.18.17",
  "$schema": "https://json.schemastore.org/package.json",
  "private": true,
  "type": "module",
@@ -1,6 +1,6 @@
 {
  "name": "@opencode-ai/console-mail",
-  "version": "1.18.16",
+  "version": "1.18.17",
  "dependencies": {
    "@jsx-email/all": "2.2.3",
    "@jsx-email/cli": "1.4.3",
@@ -1,6 +1,6 @@
 {
  "name": "@opencode-ai/console-support",
-  "version": "1.18.16",
+  "version": "1.18.17",
  "type": "module",
  "license": "MIT",
  "scripts": {
@@ -1,6 +1,6 @@
 {
  "$schema": "https://json.schemastore.org/package.json",
-  "version": "1.18.16",
+  "version": "1.18.17",
  "name": "@opencode-ai/core",
  "type": "module",
  "license": "MIT",
@@ -30,15 +30,11 @@ Guidelines:

 Complete the user's search request efficiently and report your findings clearly.`

-const PROMPT_COMPACTION = `You are an anchored context summarization assistant for coding sessions.
-
-Summarize only the conversation history you are given. The newest turns may be kept verbatim outside your summary, so focus on the older context that still matters for continuing the work.
-
-If the prompt includes a <previous-summary> block, treat it as the current anchored summary. Update it with the new history by preserving still-true details, removing stale details, and merging in new facts.
+const PROMPT_COMPACTION = `You are a context summarization agent. You are given a conversation between a user and an agent. Your goal is to produce a structured summary matching the format specified so another coding agent can continue the work.

 Always follow the exact output structure requested by the user prompt. Keep every section, preserve exact file paths and identifiers when known, and prefer terse bullets over paragraphs.

-Do not answer the conversation itself. Do not mention that you are summarizing, compacting, or merging context. Respond in the same language as the conversation.`
+Do not continue the conversation. Do not respond to any questions in the conversation. Only output the structured summary in the exact format requested by the user prompt. Respond in the same language as the conversation.`

 const PROMPT_TITLE = `You are a title generator. You output ONLY a thread title. Nothing else.

@@ -44,6 +44,15 @@ Rules:
 - Use terse bullets, not prose paragraphs.
 - Preserve exact file paths, symbols, commands, error strings, URLs, and identifiers when known.
 - Do not mention the summary process or that context was compacted.`
+const SUMMARY_UPDATE_INSTRUCTIONS = `The <prior-summary> summarizes everything that happened before the <conversation>. Construct a new summary that combines both. The <prior-summary> is discarded after this: anything you do not carry into the new summary is lost.
+
+When combining:
+- Carry forward objectives, constraints, user directives, decisions, and parallel workstreams from the <prior-summary> even when the <conversation> does not mention them. Drop only what is finished and no longer needed.
+- The <conversation> is more recent than the <prior-summary>. Where they conflict, the conversation wins: state the corrected fact and drop the old claim.
+- Add new progress, decisions, constraints, and context from the conversation.
+- Move completed work from "Active" to "Completed".
+- If a blocker has been resolved, update the summary to reflect that while keeping any details still needed to continue the work.
+- Update "Objective" and "Next Move" to reflect the current work state.`

 type Entry = {
  readonly seq: number
@@ -136,36 +145,33 @@ const select = (
  if (conversation.length === 0) return
  let total = 0
  let split = conversation.length
-  let splitPrefix = ""
-  let splitSuffix = ""
  for (let index = conversation.length - 1; index >= 0; index--) {
    const next = total + Token.estimate(conversation[index])
-    if (next > tokens) {
-      const remaining = Math.max(0, tokens - total) * 4
-      if (remaining > 0) {
-        splitPrefix = conversation[index].slice(0, -remaining)
-        splitSuffix = conversation[index].slice(-remaining)
-        split = index + 1
-      }
-      break
-    }
+    if (next > tokens) break
    total = next
    split = index
  }
  return {
-    head: [...conversation.slice(0, split), splitPrefix].filter(Boolean).join("\n\n"),
-    recent: [splitSuffix, ...conversation.slice(split)].filter(Boolean).join("\n\n"),
+    head: conversation.slice(0, split).join("\n\n"),
+    recent: conversation.slice(split).join("\n\n"),
  }
 }

-export const buildPrompt = (input: { readonly previousSummary?: string; readonly context: readonly string[] }) =>
-  [
-    input.previousSummary
-      ? `Update the anchored summary below using the conversation history above.\nPreserve still-true details, remove stale details, and merge in the new facts.\n<previous-summary>\n${input.previousSummary}\n</previous-summary>`
-      : "Create a new anchored summary from the conversation history.",
+export const buildPrompt = (input: { readonly previousSummary?: string; readonly context: readonly string[] }) => {
+  const conversation = `Here is the conversation so far:\n\n<conversation>\n${input.context.join("\n\n")}\n</conversation>`
+  if (!input.previousSummary)
+    return [
+      conversation,
+      "Create a new anchored summary from the conversation history in the <conversation> tags above so another coding agent can continue the work.",
+      SUMMARY_TEMPLATE,
+    ].join("\n\n")
+  return [
+    conversation,
+    `Here is the summary of the conversation before the <conversation> above:\n\n<prior-summary>\n${input.previousSummary}\n</prior-summary>`,
+    SUMMARY_UPDATE_INSTRUCTIONS,
    SUMMARY_TEMPLATE,
-    ...input.context,
  ].join("\n\n")
+}

 export const make = (dependencies: Dependencies) => {
  const config = settings(dependencies.config)
@@ -156,7 +156,7 @@ export const Info = Schema.Struct({
      }),
      tail_turns: Schema.optional(NonNegativeInt).annotate({
        description:
-          "Number of recent user turns, including their following assistant/tool responses, to keep verbatim during compaction (default: 2)",
+          "Maximum number of recent user turns, including their following assistant/tool responses, to keep verbatim during compaction. By default retention is limited only by the preserved token budget.",
      }),
      preserve_recent_tokens: Schema.optional(NonNegativeInt).annotate({
        description: "Maximum number of tokens from recent turns to preserve verbatim after compaction",
@@ -0,0 +1,28 @@
+import { createGroq } from "@ai-sdk/groq"
+import { expect, test } from "bun:test"
+
+test("Groq passes through unknown reasoning effort", async () => {
+  let body: Record<string, unknown> | undefined
+  const mockFetch = Object.assign(
+    async (_input: Parameters<typeof fetch>[0], init?: RequestInit) => {
+      body = JSON.parse(String(init?.body))
+      return Response.json({
+        id: "response-1",
+        created: 0,
+        model: "openai/gpt-oss-120b",
+        object: "chat.completion",
+        choices: [{ index: 0, message: { role: "assistant", content: "Hello" }, finish_reason: "stop" }],
+        usage: { prompt_tokens: 1, completion_tokens: 1, total_tokens: 2 },
+      })
+    },
+    { preconnect: fetch.preconnect },
+  )
+  const model = createGroq({ apiKey: "test", fetch: mockFetch })("openai/gpt-oss-120b")
+
+  await model.doGenerate({
+    prompt: [{ role: "user", content: [{ type: "text", text: "Hello" }] }],
+    providerOptions: { groq: { reasoningEffort: "custom" } },
+  })
+
+  expect(body?.reasoning_effort).toBe("custom")
+})
@@ -27,6 +27,32 @@ test("Mistral sends promptCacheKey as prompt_cache_key", async () => {
  expect(body?.prompt_cache_key).toBe("session-123")
 })

+test("Mistral passes through unknown reasoning effort", async () => {
+  let body: Record<string, unknown> | undefined
+  const mockFetch = Object.assign(
+    async (_input: Parameters<typeof fetch>[0], init?: RequestInit) => {
+      body = JSON.parse(String(init?.body))
+      return Response.json({
+        id: "response-1",
+        created: 0,
+        model: "mistral-large-latest",
+        object: "chat.completion",
+        choices: [{ index: 0, message: { role: "assistant", content: "Hello" }, finish_reason: "stop" }],
+        usage: { prompt_tokens: 1, completion_tokens: 1, total_tokens: 2 },
+      })
+    },
+    { preconnect: fetch.preconnect },
+  )
+  const model = createMistral({ apiKey: "test", fetch: mockFetch })("mistral-large-latest")
+
+  await model.doGenerate({
+    prompt: [{ role: "user", content: [{ type: "text", text: "Hello" }] }],
+    providerOptions: { mistral: { reasoningEffort: "custom" } },
+  })
+
+  expect(body?.reasoning_effort).toBe("custom")
+})
+
 test("Mistral round-trips native reasoning in assistant history", async () => {
  let body: { messages?: unknown[] } | undefined
  const mockFetch = Object.assign(
@@ -30,3 +30,56 @@ test("xAI Responses sends promptCacheKey as prompt_cache_key", async () => {

  expect(body?.prompt_cache_key).toBe("session-123")
 })
+
+test("xAI Responses passes through xhigh reasoning effort", async () => {
+  let body: Record<string, unknown> | undefined
+  const mockFetch = Object.assign(
+    async (_input: Parameters<typeof fetch>[0], init?: RequestInit) => {
+      body = JSON.parse(String(init?.body))
+      return Response.json({
+        id: "response-1",
+        created_at: 0,
+        model: "grok-4",
+        object: "response",
+        output: [],
+        usage: { input_tokens: 1, output_tokens: 0 },
+        status: "completed",
+      })
+    },
+    { preconnect: fetch.preconnect },
+  )
+  const model = createXai({ apiKey: "test", fetch: mockFetch }).responses("grok-4")
+
+  await model.doGenerate({
+    prompt: [{ role: "user", content: [{ type: "text", text: "Hello" }] }],
+    providerOptions: { xai: { reasoningEffort: "xhigh" } },
+  })
+
+  expect(body?.reasoning).toEqual({ effort: "xhigh" })
+})
+
+test("xAI Chat passes through xhigh reasoning effort", async () => {
+  let body: Record<string, unknown> | undefined
+  const mockFetch = Object.assign(
+    async (_input: Parameters<typeof fetch>[0], init?: RequestInit) => {
+      body = JSON.parse(String(init?.body))
+      return Response.json({
+        id: "chat-1",
+        created: 0,
+        model: "grok-4",
+        object: "chat.completion",
+        choices: [{ index: 0, message: { role: "assistant", content: "Hello" }, finish_reason: "stop" }],
+        usage: { prompt_tokens: 1, completion_tokens: 1, total_tokens: 2 },
+      })
+    },
+    { preconnect: fetch.preconnect },
+  )
+  const model = createXai({ apiKey: "test", fetch: mockFetch }).chat("grok-4")
+
+  await model.doGenerate({
+    prompt: [{ role: "user", content: [{ type: "text", text: "Hello" }] }],
+    providerOptions: { xai: { reasoningEffort: "xhigh" } },
+  })
+
+  expect(body?.reasoning_effort).toBe("xhigh")
+})
@@ -4,12 +4,32 @@ import { SessionCompaction } from "@opencode-ai/core/session/compaction"
 test("compaction prompt preserves detailed work state and relevant files", () => {
  const prompt = SessionCompaction.buildPrompt({ context: ["conversation history"] })

+  expect(prompt).toStartWith(
+    "Here is the conversation so far:\n\n<conversation>\nconversation history\n</conversation>",
+  )
+  expect(prompt.indexOf("</conversation>")).toBeLessThan(prompt.indexOf("Create a new anchored summary"))
+  expect(prompt).toContain("conversation history in the <conversation> tags above")
  expect(prompt).toContain("## Work State\n### Completed")
  expect(prompt).toContain("### Active")
  expect(prompt).toContain("### Blocked")
  expect(prompt).toContain("## Relevant Files")
 })

+test("compaction prompt gives update instructions for a prior summary", () => {
+  const prompt = SessionCompaction.buildPrompt({
+    context: ["new conversation"],
+    previousSummary: "existing summary",
+  })
+
+  expect(prompt.indexOf("<conversation>")).toBeLessThan(prompt.indexOf("<prior-summary>"))
+  expect(prompt.indexOf("</prior-summary>")).toBeLessThan(prompt.indexOf("The <prior-summary> summarizes"))
+  expect(prompt).toContain(
+    "Carry forward objectives, constraints, user directives, decisions, and parallel workstreams from the <prior-summary>",
+  )
+  expect(prompt).toContain('Move completed work from "Active" to "Completed".')
+  expect(prompt).toContain('Update "Objective" and "Next Move" to reflect the current work state.')
+})
+
 test("compaction describes tool media without embedding base64", () => {
  const base64 = "iVBORw0KGgoAAAANSUhEUgAAAAEAAAAB"
  const serialized = SessionCompaction.serializeToolContent([
@@ -1135,7 +1135,7 @@ describe("SessionRunnerLLM", () => {

      expect(requests).toHaveLength(2)
      expect(userTexts(requests[0])[0]).toContain(
-        "<previous-summary>\n## Objective\n- Preserve the task\n</previous-summary>",
+        "<prior-summary>\n## Objective\n- Preserve the task\n</prior-summary>",
      )
      expect(userTexts(requests[0])[0]).toContain("Recent exact request")
      expect((yield* (yield* SessionStore.Service).context(sessionID))[0]).toMatchObject({
@@ -1145,6 +1145,67 @@ describe("SessionRunnerLLM", () => {
    }),
  )

+  it.effect("retains only complete serialized messages during compaction", () =>
+    Effect.gen(function* () {
+      yield* setup
+      const session = yield* SessionV2.Service
+      const earlier = `EARLIER_BOUNDARY ${"a".repeat(3_000)} EARLIER_END`
+      const recent = `RECENT_BOUNDARY ${"b".repeat(3_000)} RECENT_END`
+      response = fragmentFixture("text", "text-earlier", ["Earlier answer"]).completeEvents
+      yield* session.prompt({ sessionID, prompt: Prompt.make({ text: earlier }), resume: false })
+      yield* session.resume(sessionID)
+
+      currentModel = compactModel
+      requests.length = 0
+      responses = [
+        fragmentFixture("text", "text-summary", ["## Objective\n- Preserve the task"]).completeEvents,
+        fragmentFixture("text", "text-final", ["Continued"]).completeEvents,
+      ]
+      yield* session.prompt({ sessionID, prompt: Prompt.make({ text: recent }), resume: false })
+      yield* session.resume(sessionID)
+
+      expect(requests).toHaveLength(2)
+      const summary = userTexts(requests[0])[0]
+      const continuation = userTexts(requests[1])[0]
+      expect(summary.match(/EARLIER_BOUNDARY/g)).toHaveLength(1)
+      expect(summary).toContain(`EARLIER_BOUNDARY ${"a".repeat(3_000)} EARLIER_END`)
+      expect(summary).not.toContain("RECENT_BOUNDARY")
+      expect(continuation).not.toContain("EARLIER_BOUNDARY")
+      expect(continuation).not.toContain("EARLIER_END")
+      expect(continuation).toContain("<recent-context>\n[Assistant]: Earlier answer")
+      expect(continuation).toContain(`RECENT_BOUNDARY ${"b".repeat(3_000)} RECENT_END`)
+    }),
+  )
+
+  it.effect("summarizes an oversized newest message without retaining a fragment", () =>
+    Effect.gen(function* () {
+      yield* setup
+      const session = yield* SessionV2.Service
+      response = fragmentFixture("text", "text-earlier", ["Earlier answer"]).completeEvents
+      yield* session.prompt({ sessionID, prompt: Prompt.make({ text: "Earlier question" }), resume: false })
+      yield* session.resume(sessionID)
+
+      const oversized = `OVERSIZED_BOUNDARY ${"x".repeat(4_500)} OVERSIZED_END`
+      currentModel = compactModel
+      requests.length = 0
+      responses = [
+        fragmentFixture("text", "text-summary", ["## Objective\n- Preserve the task"]).completeEvents,
+        fragmentFixture("text", "text-final", ["Continued"]).completeEvents,
+      ]
+      yield* session.prompt({ sessionID, prompt: Prompt.make({ text: oversized }), resume: false })
+      yield* session.resume(sessionID)
+
+      expect(requests).toHaveLength(2)
+      const summary = userTexts(requests[0])[0]
+      const continuation = userTexts(requests[1])[0]
+      expect(summary.match(/OVERSIZED_BOUNDARY/g)).toHaveLength(1)
+      expect(summary).toContain(oversized)
+      expect(continuation).not.toContain("OVERSIZED_BOUNDARY")
+      expect(continuation).not.toContain("OVERSIZED_END")
+      expect(continuation).toContain("<recent-context>\n\n</recent-context>")
+    }),
+  )
+
  it.effect("forces one compaction and retries after provider context overflow", () =>
    Effect.gen(function* () {
      const session = yield* setupOverflowRecovery
@@ -1,7 +1,7 @@
 {
  "name": "@opencode-ai/desktop",
  "private": true,
-  "version": "1.18.16",
+  "version": "1.18.17",
  "type": "module",
  "license": "MIT",
  "homepage": "https://opencode.ai",
@@ -1,6 +1,6 @@
 {
  "$schema": "https://json.schemastore.org/package.json",
-  "version": "1.18.16",
+  "version": "1.18.17",
  "name": "@opencode-ai/effect-drizzle-sqlite",
  "type": "module",
  "license": "MIT",
@@ -1,6 +1,6 @@
 {
  "$schema": "https://json.schemastore.org/package.json",
-  "version": "1.18.16",
+  "version": "1.18.17",
  "name": "@opencode-ai/effect-sqlite-node",
  "type": "module",
  "license": "MIT",
@@ -1,6 +1,6 @@
 {
  "name": "@opencode-ai/enterprise",
-  "version": "1.18.16",
+  "version": "1.18.17",
  "private": true,
  "type": "module",
  "license": "MIT",
@@ -1,6 +1,6 @@
 {
  "name": "@opencode-ai/function",
-  "version": "1.18.16",
+  "version": "1.18.17",
  "$schema": "https://json.schemastore.org/package.json",
  "private": true,
  "type": "module",
@@ -1,6 +1,6 @@
 {
  "$schema": "https://json.schemastore.org/package.json",
-  "version": "1.18.16",
+  "version": "1.18.17",
  "name": "@opencode-ai/http-recorder",
  "description": "Record and replay Effect HTTP client traffic with deterministic cassettes",
  "type": "module",
@@ -1,6 +1,6 @@
 {
  "$schema": "https://json.schemastore.org/package.json",
-  "version": "1.18.16",
+  "version": "1.18.17",
  "name": "@opencode-ai/llm",
  "type": "module",
  "license": "MIT",
@@ -1,6 +1,6 @@
 {
  "$schema": "https://json.schemastore.org/package.json",
-  "version": "1.18.16",
+  "version": "1.18.17",
  "name": "opencode",
  "type": "module",
  "license": "MIT",
@@ -1,9 +1,5 @@
-You are an anchored context summarization assistant for coding sessions.
-
-Summarize only the conversation history you are given. The newest turns may be kept verbatim outside your summary, so focus on the older context that still matters for continuing the work.
-
-If the prompt includes a <previous-summary> block, treat it as the current anchored summary. Update it with the new history by preserving still-true details, removing stale details, and merging in new facts.
+You are a context summarization agent. You are given a conversation between a user and an agent. Your goal is to produce a structured summary matching the format specified so another coding agent can continue the work.

 Always follow the exact output structure requested by the user prompt. Keep every section, preserve exact file paths and identifiers when known, and prefer terse bullets over paragraphs.

-Do not answer the conversation itself. Do not mention that you are summarizing, compacting, or merging context. Respond in the same language as the conversation.
+Do not continue the conversation. Do not respond to any questions in the conversation. Only output the structured summary in the exact format requested by the user prompt. Respond in the same language as the conversation.
@@ -84,6 +84,8 @@ function sdkKey(npm: string): string | undefined {
      return "gateway"
    case "@openrouter/ai-sdk-provider":
      return "openrouter"
+    case "merge-gateway-ai-sdk-provider":
+      return "mergeGateway"
    case "ai-gateway-provider":
      // ai-gateway-provider/unified wraps createOpenAICompatible({ name: "Unified" }),
      // and @ai-sdk/openai-compatible parses compatibleOptions from one of
@@ -1772,6 +1774,7 @@ function reasoningEffort(model: Provider.Model, effort: string) {
    case "@ai-sdk/togetherai":
    case "venice-ai-sdk-provider":
    case "ai-gateway-provider":
+    case "merge-gateway-ai-sdk-provider":
      return { reasoningEffort: effort }
    case "@ai-sdk/cohere":
    case "@ai-sdk/perplexity":
@@ -5,8 +5,7 @@ import { InstanceDisposed } from "@/server/event"
 import "@opencode-ai/core/account"
 import "@/server/event"
 import { Schema } from "effect"
-import { HttpApi, HttpApiEndpoint, HttpApiError, HttpApiGroup, OpenApi } from "effect/unstable/httpapi"
-import semver from "semver"
+import { HttpApi, HttpApiEndpoint, HttpApiError, HttpApiGroup, HttpApiSchema, OpenApi } from "effect/unstable/httpapi"
 import { described } from "./metadata"

 const GlobalHealth = Schema.Struct({
@@ -49,9 +48,7 @@ const GlobalEventSchema = Schema.Struct({
 }).annotate({ identifier: "GlobalEvent" })

 export const GlobalUpgradeInput = Schema.Struct({
-  target: Schema.optional(
-    Schema.String.check(Schema.makeFilter((value) => semver.valid(value) !== null || "Expected a semantic version")),
-  ),
+  target: Schema.optional(Schema.String),
 })

 const GlobalUpgradeResult = Schema.Union([
@@ -124,7 +121,7 @@ export const GlobalApi = HttpApi.make("global").add(
        }),
      ),
      HttpApiEndpoint.post("upgrade", GlobalPaths.upgrade, {
-        payload: GlobalUpgradeInput,
+        payload: [HttpApiSchema.NoContent, GlobalUpgradeInput],
        success: described(GlobalUpgradeResult, "Upgrade result"),
        error: HttpApiError.BadRequest,
      }).annotateMerge(
@@ -129,10 +129,6 @@ export const globalHandlers = HttpApiBuilder.group(RootHttpApi, "global", (handl
    const upgradeRaw = Effect.fn("GlobalHttpApi.upgradeRaw")(function* (ctx: {
      request: HttpServerRequest.HttpServerRequest
    }) {
-      const contentType = ctx.request.headers["content-type"]?.split(";", 1)[0].trim().toLowerCase()
-      if (contentType !== "application/json") {
-        return HttpServerResponse.jsonUnsafe({ success: false, error: "Expected application/json" }, { status: 415 })
-      }
      const body = yield* Effect.orDie(ctx.request.text)
      const json = parseBody(body)
      if (json === undefined) {
@@ -29,9 +29,8 @@ export const PRUNE_MINIMUM = 20_000
 export const PRUNE_PROTECT = 40_000
 const TOOL_OUTPUT_MAX_CHARS = 2_000
 const PRUNE_PROTECTED_TOOLS = ["skill"]
-const DEFAULT_TAIL_TURNS = 2
 const MIN_PRESERVE_RECENT_TOKENS = 2_000
-const MAX_PRESERVE_RECENT_TOKENS = 8_000
+const MAX_PRESERVE_RECENT_TOKENS = 15_000
 type Turn = {
  start: number
  end: number
@@ -226,27 +225,22 @@ const layer = Layer.effect(
      cfg: ConfigV1.Info
      model: Provider.Model
    }) {
-      const limit = input.cfg.compaction?.tail_turns ?? DEFAULT_TAIL_TURNS
-      if (limit <= 0) return { head: input.messages, tail_start_id: undefined }
+      const limit = input.cfg.compaction?.tail_turns
+      if (limit !== undefined && limit <= 0) return { head: input.messages, tail_start_id: undefined }
      const budget = preserveRecentBudget({ cfg: input.cfg, model: input.model })
      const all = turns(input.messages)
      if (!all.length) return { head: input.messages, tail_start_id: undefined }
-      const recent = all.slice(-limit)
-      const sizes = yield* Effect.forEach(
-        recent,
-        (turn) =>
-          estimate({
-            messages: input.messages.slice(turn.start, turn.end),
-            model: input.model,
-          }),
-        { concurrency: 1 },
-      )
+      const recent = limit === undefined ? all : all.slice(-limit)

      let total = 0
      let keep: Tail | undefined
      for (let i = recent.length - 1; i >= 0; i--) {
        const turn = recent[i]!
-        const size = sizes[i]
+        // estimate lazily so cost stays proportional to the retained tail, not the whole session
+        const size = yield* estimate({
+          messages: input.messages.slice(turn.start, turn.end),
+          model: input.model,
+        })
        if (total + size <= budget) {
          total += size
          keep = { start: turn.start, id: turn.id }
@@ -381,10 +375,20 @@ const layer = Layer.effect(
        { sessionID: input.sessionID },
        { context: [], prompt: undefined },
      )
-      const nextPrompt = compacting.prompt ?? buildPrompt({ previousSummary, context: compacting.context })
      const msgs = structuredClone(selected.head)
      yield* plugin.trigger("experimental.chat.messages.transform", {}, { messages: msgs })
      const conversation = msgs.map(serialize).filter(Boolean).join("\n\n")
+      const nextPrompt =
+        compacting.prompt ??
+        [
+          buildPrompt({
+            previousSummary,
+            context: [conversation],
+          }),
+          ...compacting.context,
+        ]
+          .filter(Boolean)
+          .join("\n\n")
      const ctx = yield* InstanceState.context
      const msg: SessionV1.Assistant = {
        id: MessageID.ascending(),
@@ -430,7 +434,10 @@ const layer = Layer.effect(
            content: [
              {
                type: "text",
-                text: [nextPrompt, "The following is the conversation history:", conversation]
+                text: [
+                  nextPrompt,
+                  ...(compacting.prompt ? ["The following is the conversation history:", conversation] : []),
+                ]
                  .filter(Boolean)
                  .join("\n\n"),
              },
@@ -40,7 +40,11 @@ export function provider(model: Provider.Model) {
  if (model.api.id.includes("gemini-")) return [PROMPT_GEMINI]
  if (model.api.id.includes("claude")) return [PROMPT_ANTHROPIC]
  if (model.api.id.toLowerCase().includes("trinity")) return [PROMPT_TRINITY]
-  if (model.api.id.toLowerCase().includes("kimi")) return [PROMPT_KIMI]
+  if (
+    model.api.id.toLowerCase().includes("kimi") ||
+    ["kimi-for-coding", "moonshotai", "moonshotai-cn"].includes(model.providerID)
+  )
+    return [PROMPT_KIMI]
  return [PROMPT_DEFAULT]
 }

@@ -1548,6 +1548,33 @@ test("models.dev reasoning options replace generated variants and unsupported to
  expect(models["gemini-3-pro-fast"].variants).toEqual(models.override.variants)
 })

+test("MERGE Gateway exposes declared effort variants without model-specific handling", () => {
+  const provider = {
+    id: "merge-gateway",
+    name: "MERGE Gateway",
+    env: ["MERGE_GATEWAY_API_KEY"],
+    npm: "merge-gateway-ai-sdk-provider",
+    models: {
+      "openai/gpt-5.6-sol": {
+        id: "openai/gpt-5.6-sol",
+        name: "GPT-5.6 Sol",
+        reasoning: true,
+        reasoning_options: [{ type: "effort", values: ["none", "low", "medium", "high", "xhigh", "max"] }],
+        limit: { context: 128_000, output: 64_000 },
+      },
+    },
+  } as unknown as ModelsDev.Provider
+
+  expect(Provider.fromModelsDevProvider(provider).models["openai/gpt-5.6-sol"].variants).toEqual({
+    none: { reasoningEffort: "none" },
+    low: { reasoningEffort: "low" },
+    medium: { reasoningEffort: "medium" },
+    high: { reasoningEffort: "high" },
+    xhigh: { reasoningEffort: "xhigh" },
+    max: { reasoningEffort: "max" },
+  })
+})
+
 test("public provider info omits invalid models", () => {
  const provider = Provider.fromModelsDevProvider({
    id: "test",
@@ -3370,6 +3370,7 @@ describe("ProviderTransform.reasoningVariants", () => {
    ["@ai-sdk/togetherai", { reasoningEffort: "high" }],
    ["venice-ai-sdk-provider", { reasoningEffort: "high" }],
    ["ai-gateway-provider", { reasoningEffort: "high" }],
+    ["merge-gateway-ai-sdk-provider", { reasoningEffort: "high" }],
    ["@ai-sdk/amazon-bedrock", { reasoningConfig: { type: "enabled", maxReasoningEffort: "high" } }],
  ])("converts effort for %s", (npm, expected, ...args) => {
    const id = args[0] as string | undefined
@@ -5555,6 +5556,25 @@ describe("ProviderTransform.providerOptions - ai-gateway-provider", () => {
  })
 })

+describe("ProviderTransform.providerOptions - merge-gateway-ai-sdk-provider", () => {
+  const model = {
+    id: "merge-gateway/openai/gpt-5.6-sol",
+    providerID: "merge-gateway",
+    api: {
+      id: "openai/gpt-5.6-sol",
+      url: "https://api-gateway.merge.dev/v1/ai-sdk",
+      npm: "merge-gateway-ai-sdk-provider",
+    },
+    capabilities: { reasoning: true },
+  } as any
+
+  test("routes normalized effort under the adapter's mergeGateway key", () => {
+    expect(ProviderTransform.providerOptions(model, { reasoningEffort: "high" })).toEqual({
+      mergeGateway: { reasoningEffort: "high" },
+    })
+  })
+})
+
 describe("ProviderTransform.options - kimi family adaptive thinking", () => {
  const createModel = (overrides: Record<string, any> = {}) =>
    ({
@@ -43,12 +43,9 @@ const apiLayer = HttpRouter.serve(
 const it = testEffect(apiLayer)

 describe("global HttpApi", () => {
-  it.live("upgrades to latest with an empty JSON payload", () =>
+  it.live("upgrades to latest when the request body is omitted", () =>
    Effect.gen(function* () {
-      const response = yield* HttpClientRequest.post(GlobalPaths.upgrade).pipe(
-        HttpClientRequest.bodyJson({}),
-        Effect.flatMap(HttpClient.execute),
-      )
+      const response = yield* HttpClient.post(GlobalPaths.upgrade)

      expect(response.status).toBe(200)
      expect(yield* response.json).toEqual({ success: true, version: "9.9.9" })
@@ -66,40 +63,4 @@ describe("global HttpApi", () => {
      expect(yield* response.json).toEqual({ success: false, error: "Invalid request body" })
    }),
  )
-
-  it.live("rejects non-JSON upgrade payloads", () =>
-    Effect.gen(function* () {
-      const response = yield* HttpClientRequest.post(GlobalPaths.upgrade).pipe(
-        HttpClientRequest.setBody(HttpBody.text('{"target":"1.2.3"}', "text/plain")),
-        HttpClient.execute,
-      )
-
-      expect(response.status).toBe(415)
-      expect(yield* response.json).toEqual({ success: false, error: "Expected application/json" })
-    }),
-  )
-
-  it.live("rejects non-semver upgrade targets", () =>
-    Effect.gen(function* () {
-      const response = yield* HttpClientRequest.post(GlobalPaths.upgrade).pipe(
-        HttpClientRequest.bodyJson({ target: "latest" }),
-        Effect.flatMap(HttpClient.execute),
-      )
-
-      expect(response.status).toBe(400)
-      expect(yield* response.json).toEqual({ success: false, error: "Invalid request body" })
-    }),
-  )
-
-  it.live("accepts semantic version upgrade targets", () =>
-    Effect.gen(function* () {
-      const response = yield* HttpClientRequest.post(GlobalPaths.upgrade).pipe(
-        HttpClientRequest.bodyJson({ target: "1.2.3-beta.1" }),
-        Effect.flatMap(HttpClient.execute),
-      )
-
-      expect(response.status).toBe(200)
-      expect(yield* response.json).toEqual({ success: true, version: "1.2.3-beta.1" })
-    }),
-  )
 })
@@ -365,6 +365,20 @@ function autocontinue(enabled: boolean) {
  })
 }

+function compactionContext(context: string) {
+  return Layer.mock(Plugin.Service)({
+    trigger: <Name extends string, Input, Output>(name: Name, _input: Input, output: Output) => {
+      if (name !== "experimental.session.compacting") return Effect.succeed(output)
+      return Effect.sync(() => {
+        ;(output as { context: string[] }).context.push(context)
+        return output
+      })
+    },
+    list: () => Effect.succeed([]),
+    init: () => Effect.void,
+  })
+}
+
 describe("session.compaction.isOverflow", () => {
  it.live(
    "returns true when token count exceeds usable context",
@@ -1389,11 +1403,21 @@ describe("session.compaction.process", () => {
        const captured = JSON.stringify(messages)
        expect(messages).toHaveLength(1)
        expect(messages[0]?.role).toBe("user")
+        expect(captured).toContain("Here is the conversation so far:")
+        expect(captured).toContain("<conversation>")
+        expect(captured.indexOf("[User]: older context")).toBeLessThan(
+          captured.indexOf("Create a new anchored summary"),
+        )
        expect(captured).toContain("[User]: older context")
        expect(captured).not.toContain("keep this turn")
        expect(captured).not.toContain("and this one too")
        expect(captured).not.toContain("What did we do so far?")
-      }).pipe(withCompaction({ llm: stub.llmLayer }))
+      }).pipe(
+        withCompaction({
+          llm: stub.llmLayer,
+          config: cfg({ tail_turns: 2, preserve_recent_tokens: 10_000 }),
+        }),
+      )
    },
    { git: true },
  )
@@ -1430,9 +1454,11 @@ describe("session.compaction.process", () => {
        expect(parent).toBeTruthy()
        yield* SessionCompaction.use.process({ parentID: parent!, messages: msgs, sessionID: session.id, auto: false })

-        expect(captured).toContain("<previous-summary>")
+        expect(captured).toContain("<prior-summary>")
        expect(captured).toContain("summary one")
        expect(captured.match(/summary one/g)?.length).toBe(1)
+        expect(captured.indexOf("latest turn")).toBeLessThan(captured.indexOf("<prior-summary>"))
+        expect(captured).toContain("summary of the conversation before the <conversation> above")
        expect(captured).toContain("## Important Details")
        expect(captured).toContain("## Work State")
      }).pipe(withCompaction({ llm: stub.llmLayer }))
@@ -1440,6 +1466,49 @@ describe("session.compaction.process", () => {
    { git: true },
  )

+  itCompaction.instance(
+    "keeps plugin context outside the serialized conversation",
+    () => {
+      const stub = llm()
+      let captured = ""
+      stub.push(
+        reply("summary", (input) => {
+          captured = JSON.stringify(input.messages)
+        }),
+      )
+
+      return Effect.gen(function* () {
+        const ssn = yield* SessionNs.Service
+        const session = yield* ssn.create({})
+        yield* createUserMessage(session.id, "older context")
+        yield* createUserMessage(session.id, "keep this turn")
+        yield* createUserMessage(session.id, "and this one too")
+        yield* createCompactionMarker(session.id)
+
+        const msgs = yield* ssn.messages({ sessionID: session.id })
+        const parent = msgs.at(-1)?.info.id
+        expect(parent).toBeTruthy()
+        yield* SessionCompaction.use.process({
+          parentID: parent!,
+          messages: msgs,
+          sessionID: session.id,
+          auto: false,
+        })
+
+        expect(captured).toContain("Prioritize unresolved migration details")
+        expect(captured.indexOf("</conversation>")).toBeLessThan(
+          captured.indexOf("Prioritize unresolved migration details"),
+        )
+      }).pipe(
+        withCompaction({
+          llm: stub.llmLayer,
+          plugin: compactionContext("Prioritize unresolved migration details"),
+        }),
+      )
+    },
+    { git: true },
+  )
+
  itCompaction.instance(
    "serializes repeated compaction history as one user message",
    () => {
@@ -102,6 +102,13 @@ describe("session.system", () => {
    }
  })

+  test("selects the Kimi prompt for official provider model IDs", () => {
+    for (const providerID of ["kimi-for-coding", "moonshotai", "moonshotai-cn"]) {
+      const prompt = SystemPrompt.provider({ providerID, api: { id: "k3" } } as Provider.Model)[0]
+      expect(prompt).toContain("# Prompt and Tool Use")
+    }
+  })
+
  it.effect("skills output is sorted by name and stable across calls", () =>
    Effect.gen(function* () {
      const prompt = yield* SystemPrompt.Service
@@ -1,7 +1,7 @@
 {
  "$schema": "https://json.schemastore.org/package.json",
  "name": "@opencode-ai/plugin",
-  "version": "1.18.16",
+  "version": "1.18.17",
  "type": "module",
  "license": "MIT",
  "scripts": {
@@ -1,7 +1,7 @@
 {
  "$schema": "https://json.schemastore.org/package.json",
  "name": "@opencode-ai/sdk",
-  "version": "1.18.16",
+  "version": "1.18.17",
  "type": "module",
  "license": "MIT",
  "scripts": {
@@ -1,7 +1,7 @@
 {
  "$schema": "https://json.schemastore.org/package.json",
  "name": "@opencode-ai/server",
-  "version": "1.18.16",
+  "version": "1.18.17",
  "private": true,
  "type": "module",
  "license": "MIT",
@@ -1,6 +1,6 @@
 {
  "name": "@opencode-ai/session-ui",
-  "version": "1.18.16",
+  "version": "1.18.17",
  "private": true,
  "type": "module",
  "license": "MIT",
@@ -1,6 +1,6 @@
 {
  "name": "@opencode-ai/slack",
-  "version": "1.18.16",
+  "version": "1.18.17",
  "type": "module",
  "license": "MIT",
  "scripts": {
@@ -1,7 +1,7 @@
 {
  "$schema": "https://json.schemastore.org/package.json",
  "name": "@opencode-ai/stats-app",
-  "version": "1.18.16",
+  "version": "1.18.17",
  "private": true,
  "type": "module",
  "license": "MIT",
@@ -1,7 +1,7 @@
 {
  "$schema": "https://json.schemastore.org/package.json",
  "name": "@opencode-ai/stats-core",
-  "version": "1.18.16",
+  "version": "1.18.17",
  "private": true,
  "type": "module",
  "license": "MIT",
@@ -12,6 +12,7 @@
    "./database": "./src/database.ts",
    "./database/*": "./src/database/*.ts",
    "./domain/*": "./src/domain/*.ts",
+    "./r2-sql": "./src/r2-sql.ts",
    "./runtime": "./src/runtime.ts",
    "./stat-sync": "./src/stat-sync.ts"
  },
@@ -1,5 +1,5 @@
 import { describe, expect, test } from "bun:test"
-import { toGeoAggregate, toModelAggregate, toProviderAggregate } from "./inference"
+import { buildStatsQueries, toGeoAggregate, toModelAggregate, toProviderAggregate } from "./inference"
 import { modelAuthor, normalizeInferenceModel, statModel, statProvider } from "./model-normalization"

 describe("inference stat normalization", () => {
@@ -82,6 +82,55 @@ describe("inference stat normalization", () => {
      }),
    ).toMatchObject([{ period_key: "2026-W20" }])
  })
+
+  test("builds bounded R2 SQL queries for each day and week", () => {
+    const queries = buildStatsQueries(new Date("2026-08-10T00:00:00.000Z"), new Date("2026-08-12T12:00:00.000Z"), {
+      namespace: "inference",
+      table: "generation",
+      dataset: "zen",
+    })
+
+    expect(queries).toHaveLength(8)
+    expect(queries[0]).toContain("'week' AS grain")
+    expect(queries[0]).toContain("'2026-W33' AS period_key")
+    expect(queries[2]).toContain("'2026-08-10' AS period_key")
+    expect(queries[6]).toContain("'2026-08-12' AS period_key")
+    expect(queries[0]).toContain('FROM "inference"."generation"')
+    expect(queries[0]).toContain("event_type = 'generation.completed'")
+    expect(queries[0]).toContain("product = 'go'")
+    expect(queries[0]).toContain("LIMIT 10000")
+    expect(queries[0]).toContain("approx_distinct(session) AS sessions")
+    expect(queries[1]).toContain("'geo_model' ELSE 'geo'")
+    expect(queries[1]).toContain("0 AS sessions")
+  })
+
+  test("aligns periods to UTC calendar boundaries", () => {
+    const queries = buildStatsQueries(new Date("2026-06-17T15:56:00.000Z"), new Date("2026-06-19T15:56:00.000Z"), {
+      namespace: "inference",
+      table: "generation",
+      dataset: "zen",
+    })
+
+    expect(queries).toHaveLength(8)
+    expect(queries[0]).toContain("'2026-W25' AS period_key")
+    expect(queries[0]).toContain("started_at >= '2026-06-15T00:00:00.000Z'")
+    expect(queries[2]).toContain("'2026-06-17' AS period_key")
+    expect(queries[2]).toContain("started_at >= '2026-06-17T00:00:00.000Z'")
+    expect(queries[2]).toContain("started_at < '2026-06-18T00:00:00.000Z'")
+    expect(queries[6]).toContain("'2026-06-19' AS period_key")
+    expect(queries[6]).toContain("started_at < '2026-06-19T15:56:00.000Z'")
+  })
+
+  test("uses an exclusive live and legacy source handoff", () => {
+    const [query] = buildStatsQueries(new Date("2026-08-11T00:00:00.000Z"), new Date("2026-08-12T00:00:00.000Z"), {
+      namespace: "inference",
+      table: "generation",
+      dataset: "zen",
+    })
+
+    expect(query).toContain("(source = 'inference-legacy' AND started_at < '2026-08-11T10:57:48.186Z')")
+    expect(query).toContain("(source = 'inference' AND started_at >= '2026-08-11T10:57:48.186Z')")
+  })
 })

 function aggregate(model: string, provider: string) {
@@ -1,5 +1,5 @@
 import { Resource } from "sst/resource"
-import type { AthenaData } from "../athena"
+import type { R2SqlData } from "../r2-sql"
 import type { GeoStatAggregate } from "./geo"
 import type { ModelStatAggregate } from "./model"
 import {
@@ -10,25 +10,80 @@ import {
  statProvider,
 } from "./model-normalization"
 import type { ProviderStatAggregate } from "./provider"
-import { normalizeCountry, normalizeTier, type StatBaseAggregate } from "./stat"
+import {
+  normalizeCountry,
+  normalizeTier,
+  periodKeyFor,
+  startOfIsoWeek,
+  startOfUtcDay,
+  type StatBaseAggregate,
+} from "./stat"

 export type StatDimension = "model" | "provider" | "geo" | "geo_model"
+export type StatsQuerySource = { namespace: string; table: string; dataset: string }
+type StatsQueryFamily = "usage" | "geo"

-// All stat dimensions and both grains are computed in one query via GROUPING SETS so
-// the source table is scanned once per sync pass; separate queries per dimension (and
-// the previous weekly/daily UNION ALL) each re-scanned the same events.
-export function buildStatsQuery(periodStart: Date, periodEnd: Date) {
-  const periodStartValue = sqlString(periodStart.toISOString())
-  const periodEndValue = sqlString(periodEnd.toISOString())
-  const periodStartDateValue = sqlString(periodStart.toISOString().slice(0, 10))
-  const periodEndDateValue = sqlString(periodEnd.toISOString().slice(0, 10))
-  const sourceTable = [Resource.InferenceEvent.catalog, Resource.InferenceEvent.database, Resource.InferenceEvent.table]
-    .map(sqlIdentifier)
-    .join(".")
+const DAY_MS = 86_400_000
+const WEEK_MS = 7 * DAY_MS
+// The typed production stream began before the legacy backfill's original end
+// boundary. Use one exclusive handoff so the overlapping rows are never counted
+// from both sources.
+const LIVE_SOURCE_START = "2026-08-11T10:57:48.186Z"
+
+// R2 SQL limits result sets to 10,000 rows and does not support OFFSET. Two
+// queries per day/week keep each result bounded and avoid combining the costly
+// distinct user/session aggregates with the high-cardinality geo dimensions.
+export function buildStatsQueries(periodStart: Date, periodEnd: Date, input?: StatsQuerySource) {
+  const source = input ?? {
+    namespace: Resource.R2Sql.namespace,
+    table: Resource.R2Sql.table,
+    dataset: Resource.StatsSyncConfig.dataset,
+  }
+  return [...statPeriods("week", periodStart, periodEnd), ...statPeriods("day", periodStart, periodEnd)].flatMap(
+    (period) => [buildStatsQuery(period, source, "usage"), buildStatsQuery(period, source, "geo")],
+  )
+}
+
+function buildStatsQuery(
+  period: { grain: "day" | "week"; key: string; start: Date; end: Date },
+  source: StatsQuerySource,
+  family: StatsQueryFamily,
+) {
+  const periodStartValue = sqlString(period.start.toISOString())
+  const periodEndValue = sqlString(period.end.toISOString())
+  const ingestEndValue = sqlString(new Date(period.end.getTime() + DAY_MS).toISOString())
+  const sourceTable = [source.namespace, source.table].map(sqlIdentifier).join(".")
+  const dimensions =
+    family === "usage"
+      ? `CASE WHEN grouping(model) = 0 THEN 'model' ELSE 'provider' END AS dimension,
+  tier,
+  provider,
+  CASE WHEN grouping(model) = 0 THEN model END AS model,
+  CASE WHEN grouping(model) = 0 THEN COALESCE(MAX(NULLIF(provider_model, '')), '') END AS provider_model,
+  null AS country,
+  null AS continent`
+      : `CASE WHEN grouping(model) = 0 THEN 'geo_model' ELSE 'geo' END AS dimension,
+  tier,
+  CASE WHEN grouping(model) = 0 THEN provider ELSE 'all' END AS provider,
+  CASE WHEN grouping(model) = 0 THEN model ELSE 'all' END AS model,
+  null AS provider_model,
+  country,
+  COALESCE(MAX(NULLIF(continent, '')), '') AS continent`
+  const distinctColumns =
+    family === "usage"
+      ? `approx_distinct(session) AS sessions,
+    approx_distinct(user_key) AS unique_users`
+      : `0 AS sessions,
+    0 AS unique_users`
+  const groupingSets =
+    family === "usage"
+      ? `(tier, provider, model),
+  (tier, provider)`
+      : `(tier, country),
+  (tier, provider, model, country)`
  const aggregateColumns = `
-    COUNT(DISTINCT session) AS sessions,
+    ${distinctColumns},
    COUNT(*) AS requests,
-    COUNT(DISTINCT user_key) AS unique_users,
    COALESCE(SUM(tokens_input), 0) AS input_tokens,
    COALESCE(SUM(tokens_output), 0) AS output_tokens,
    COALESCE(SUM(tokens_reasoning), 0) AS reasoning_tokens,
@@ -38,65 +93,61 @@ export function buildStatsQuery(periodStart: Date, periodEnd: Date) {
    COALESCE(SUM(cost_output_microcents), 0) AS output_cost_microcents,
    COALESCE(SUM(cost_total_microcents), 0) AS total_cost_microcents,
    AVG(duration_ms) AS avg_duration_ms,
-    approx_percentile(CAST(duration_ms AS double), 0.5) AS p50_duration_ms,
-    approx_percentile(CAST(duration_ms AS double), 0.95) AS p95_duration_ms,
+    null AS p50_duration_ms,
+    null AS p95_duration_ms,
    AVG(ttfb_ms) AS avg_ttfb_ms,
-    approx_percentile(CAST(ttfb_ms AS double), 0.5) AS p50_ttfb_ms,
-    approx_percentile(CAST(ttfb_ms AS double), 0.95) AS p95_ttfb_ms,
+    null AS p50_ttfb_ms,
+    null AS p95_ttfb_ms,
    AVG(output_tps) AS avg_output_tps,
-    SUM(CASE WHEN status >= 200 AND status < 400 THEN 1 ELSE 0 END) AS success_count,
-    SUM(CASE WHEN status >= 400 THEN 1 ELSE 0 END) AS error_count,
+    SUM(CASE WHEN outcome = 'succeeded' THEN 1 ELSE 0 END) AS success_count,
+    SUM(CASE WHEN outcome = 'failed' THEN 1 ELSE 0 END) AS error_count,
    COUNT(*) AS sample_count`

  return `
 WITH normalized AS (
  SELECT
-    from_iso8601_timestamp(event_timestamp) AS event_time,
-    model AS raw_model,
-    ${statModelSql("model", "provider_model")} AS model,
-    COALESCE(NULLIF(provider_model, ''), '') AS provider_model,
-    COALESCE(NULLIF(provider, ''), '') AS raw_provider,
-    UPPER(COALESCE(NULLIF(cf_country, ''), 'ZZ')) AS country,
-    COALESCE(NULLIF(cf_continent, ''), '') AS continent,
-    session,
-    COALESCE(NULLIF(workspace, ''), '') AS workspace,
-    COALESCE(NULLIF(api_key, ''), '') AS api_key,
+    model_requested AS raw_model,
+    ${statModelSql("model_requested", "route_model")} AS model,
+    COALESCE(NULLIF(route_model, ''), '') AS provider_model,
+    COALESCE(NULLIF(provider_id, ''), '') AS raw_provider,
+    UPPER(COALESCE(NULLIF(country, ''), 'ZZ')) AS country,
+    COALESCE(NULLIF(continent, ''), '') AS continent,
+    session_id AS session,
+    COALESCE(NULLIF(workspace_id, ''), '') AS workspace,
+    COALESCE(NULLIF(service_api_key_id, ''), '') AS api_key,
    COALESCE(NULLIF(user_id, ''), '') AS user_id,
-    status,
-    duration AS duration_ms,
-    time_to_first_byte AS ttfb_ms,
-    timestamp_first_byte,
-    timestamp_last_byte,
+    outcome,
+    duration_ms,
+    time_to_first_token_ms AS ttfb_ms,
+    CASE
+      WHEN first_token_at IS NULL OR last_token_at IS NULL THEN null
+      ELSE date_part('epoch', last_token_at) - date_part('epoch', first_token_at)
+    END AS output_seconds,
    tokens_input,
    tokens_output,
    tokens_reasoning,
    tokens_cache_read,
-    tokens_cache_write_5m,
-    tokens_cache_write_1h,
-    cost_input_microcents,
-    cost_output_microcents,
-    cost_total_microcents,
-    cost_input,
-    cost_output,
-    cost_total,
-    source
+    tokens_cache_write,
+    cost_input AS cost_input_microcents,
+    cost_output AS cost_output_microcents,
+    cost_total AS cost_total_microcents
  FROM ${sourceTable}
-  WHERE event_type = 'completions'
-    AND model IS NOT NULL
-    AND model <> ''
-    AND source = 'lite'
-    AND event_date >= ${periodStartDateValue}
-    AND event_date <= ${periodEndDateValue}
-    AND event_timestamp >= ${periodStartValue}
-    AND event_timestamp < ${periodEndValue}
+  WHERE event_type = 'generation.completed'
+    AND source IN ('inference', 'inference-legacy')
+    AND (
+      (source = 'inference-legacy' AND started_at < ${sqlString(LIVE_SOURCE_START)})
+      OR (source = 'inference' AND started_at >= ${sqlString(LIVE_SOURCE_START)})
+    )
+    AND product = 'go'
+    AND model_requested IS NOT NULL
+    AND model_requested <> ''
+    AND __ingest_ts >= ${periodStartValue}
+    AND __ingest_ts < ${ingestEndValue}
+    AND started_at >= ${periodStartValue}
+    AND started_at < ${periodEndValue}
 ), filtered AS (
  SELECT
-    event_time,
-    CASE
-      WHEN source = 'lite' THEN 'Go'
-      WHEN raw_model IN ('gpt-5-nano', 'grok-code', 'big-pickle') OR regexp_like(raw_model, '-free(:global)?$') THEN 'Free'
-      ELSE 'Paid'
-    END AS tier,
+    'Go' AS tier,
    ${statProviderSql("model", "provider_model", "raw_provider")} AS provider,
    provider_model,
    model,
@@ -104,63 +155,39 @@ WITH normalized AS (
    continent,
    session,
    COALESCE(NULLIF(user_id, ''), NULLIF(workspace, ''), NULLIF(api_key, '')) AS user_key,
-    status,
+    outcome,
    duration_ms,
    ttfb_ms,
    CASE
-      WHEN timestamp_last_byte - timestamp_first_byte < 100 THEN null
-      ELSE CAST(tokens_output AS double) / (timestamp_last_byte - timestamp_first_byte) * 1000
+      WHEN output_seconds < 0.1 THEN null
+      ELSE CAST(tokens_output AS double) / output_seconds
    END AS output_tps,
    tokens_input,
    tokens_output,
    tokens_reasoning,
    tokens_cache_read,
-    COALESCE(tokens_cache_read, 0) + COALESCE(tokens_cache_write_5m, 0) + COALESCE(tokens_cache_write_1h, 0) + COALESCE(tokens_input, 0) + COALESCE(tokens_output, 0) AS tokens_total,
-    COALESCE(cost_input_microcents, cost_input * 1000000) AS cost_input_microcents,
-    COALESCE(cost_output_microcents, cost_output * 1000000) AS cost_output_microcents,
-    COALESCE(cost_total_microcents, cost_total * 1000000) AS cost_total_microcents
+    COALESCE(tokens_cache_read, 0) + COALESCE(tokens_cache_write, 0) + COALESCE(tokens_input, 0) + COALESCE(tokens_output, 0) AS tokens_total,
+    cost_input_microcents,
+    cost_output_microcents,
+    cost_total_microcents
  FROM normalized
  WHERE lower(model) NOT IN (${[...EXCLUDED_MODELS].map(sqlString).join(", ")})
-), periods AS (
-  SELECT
-    concat(CAST(year_of_week(event_time) AS varchar), '-W', lpad(CAST(week(event_time) AS varchar), 2, '0')) AS week_key,
-    substr(to_iso8601(date_trunc('day', event_time)), 1, 10) AS day_key,
-    *
-  FROM filtered
 )
 SELECT
-  CASE WHEN grouping(week_key) = 0 THEN 'week' ELSE 'day' END AS grain,
-  COALESCE(week_key, day_key) AS period_key,
-  ${sqlString(Resource.StatsSyncConfig.dataset)} AS dataset,
-  CASE
-    WHEN grouping(country) = 0 AND grouping(model) = 0 THEN 'geo_model'
-    WHEN grouping(country) = 0 THEN 'geo'
-    WHEN grouping(model) = 0 THEN 'model'
-    ELSE 'provider'
-  END AS dimension,
-  tier,
-  CASE WHEN grouping(provider) = 0 THEN provider ELSE 'all' END AS provider,
-  CASE WHEN grouping(model) = 0 THEN model WHEN grouping(country) = 0 THEN 'all' END AS model,
-  CASE WHEN grouping(model) = 0 AND grouping(country) = 1 THEN COALESCE(MAX(NULLIF(provider_model, '')), '') END AS provider_model,
-  CASE WHEN grouping(country) = 0 THEN country END AS country,
-  CASE WHEN grouping(country) = 0 THEN COALESCE(MAX(NULLIF(continent, '')), '') END AS continent,
+  ${sqlString(period.grain)} AS grain,
+  ${sqlString(period.key)} AS period_key,
+  ${sqlString(source.dataset)} AS dataset,
+  ${dimensions},
  ${aggregateColumns}
-FROM periods
+FROM filtered
 GROUP BY GROUPING SETS (
-  (week_key, tier, provider, model),
-  (week_key, tier, provider),
-  (week_key, tier, country),
-  (week_key, tier, provider, model, country),
-  (day_key, tier, provider, model),
-  (day_key, tier, provider),
-  (day_key, tier, country),
-  (day_key, tier, provider, model, country)
+  ${groupingSets}
 )
-ORDER BY grain, period_key, total_tokens DESC
+LIMIT 10000
 `
 }

-export function toModelAggregate(data: AthenaData): ModelStatAggregate[] {
+export function toModelAggregate(data: R2SqlData): ModelStatAggregate[] {
  const model = statModel(data.model, data.provider_model)
  const provider = statProvider(model, data.provider_model, data.provider)
  if (!provider) return []
@@ -170,13 +197,13 @@ export function toModelAggregate(data: AthenaData): ModelStatAggregate[] {
  ])
 }

-export function toProviderAggregate(data: AthenaData): ProviderStatAggregate[] {
+export function toProviderAggregate(data: R2SqlData): ProviderStatAggregate[] {
  return toStatBaseAggregate(data).flatMap((base) => [
    { ...base, provider: statProvider(data.model, data.provider_model, data.provider) || "unknown" },
  ])
 }

-export function toGeoAggregate(data: AthenaData): GeoStatAggregate[] {
+export function toGeoAggregate(data: R2SqlData): GeoStatAggregate[] {
  return toStatBaseAggregate(data).flatMap((base) => [
    {
      ...base,
@@ -188,7 +215,7 @@ export function toGeoAggregate(data: AthenaData): GeoStatAggregate[] {
  ])
 }

-function toStatBaseAggregate(data: AthenaData): StatBaseAggregate[] {
+function toStatBaseAggregate(data: R2SqlData): StatBaseAggregate[] {
  const grain = data.grain === "day" || data.grain === "week" ? data.grain : undefined
  if (!grain || !data.period_key) return []

@@ -223,21 +250,21 @@ function toStatBaseAggregate(data: AthenaData): StatBaseAggregate[] {
  ]
 }

-function integer(data: AthenaData, key: string) {
+function integer(data: R2SqlData, key: string) {
  return Math.round(number(data, key))
 }

-function nullableNumber(data: AthenaData, key: string) {
+function nullableNumber(data: R2SqlData, key: string) {
  if (data[key] === undefined || data[key] === "") return null
  return Number(number(data, key).toFixed(2))
 }

-function nullableInteger(data: AthenaData, key: string) {
+function nullableInteger(data: R2SqlData, key: string) {
  if (data[key] === undefined || data[key] === "") return null
  return Math.round(number(data, key))
 }

-function number(data: AthenaData, key: string) {
+function number(data: R2SqlData, key: string) {
  const value = Number(data[key])
  return Number.isFinite(value) ? value : 0
 }
@@ -250,6 +277,21 @@ function sqlString(value: string) {
  return `'${value.replace(/'/g, "''")}'`
 }

+function statPeriods(grain: "day" | "week", periodStart: Date, periodEnd: Date) {
+  const interval = grain === "day" ? DAY_MS : WEEK_MS
+  const first = grain === "day" ? startOfUtcDay(periodStart) : startOfIsoWeek(periodStart)
+  const count = Math.max(0, Math.ceil((periodEnd.getTime() - first.getTime()) / interval))
+  return Array.from({ length: count }, (_, index) => {
+    const start = new Date(first.getTime() + index * interval)
+    return {
+      grain,
+      key: periodKeyFor(grain, start),
+      start,
+      end: new Date(Math.min(start.getTime() + interval, periodEnd.getTime())),
+    }
+  })
+}
+
 function statModelSql(model: string, providerModel: string) {
  return `COALESCE(NULLIF(regexp_replace(CASE
      WHEN lower(${model}) = 'big-pickle' THEN NULLIF(${providerModel}, '')
@@ -0,0 +1,105 @@
+import { Context, Effect, Layer, Schema } from "effect"
+import { Resource } from "sst/resource"
+
+const R2_SQL_MAX_ROWS = 10_000
+const R2SqlValue = Schema.Union([Schema.String, Schema.Number, Schema.Boolean, Schema.Null])
+const R2SqlResponse = Schema.Struct({
+  success: Schema.Boolean,
+  result: Schema.optional(
+    Schema.NullOr(
+      Schema.Struct({
+        request_id: Schema.String,
+        rows: Schema.Array(Schema.Record(Schema.String, R2SqlValue)),
+      }),
+    ),
+  ),
+  errors: Schema.Array(Schema.Unknown),
+})
+const decodeResponse = Schema.decodeUnknownEffect(Schema.fromJsonString(R2SqlResponse))
+
+export type R2SqlData = Record<string, string>
+
+export class R2SqlQueryError extends Error {
+  readonly _tag = "R2SqlQueryError"
+  readonly requestId?: string
+  readonly status?: number
+
+  constructor(input: { message: string; requestId?: string; status?: number; cause?: unknown }) {
+    super(input.message, { cause: input.cause })
+    this.name = "R2SqlQueryError"
+    this.requestId = input.requestId
+    this.status = input.status
+  }
+}
+
+export declare namespace R2Sql {
+  export interface Service {
+    readonly query: (query: string) => Effect.Effect<R2SqlData[], R2SqlQueryError>
+  }
+}
+
+export class R2Sql extends Context.Service<R2Sql, R2Sql.Service>()("@opencode/stats/R2Sql") {
+  static readonly layer: Layer.Layer<R2Sql> = Layer.succeed(
+    R2Sql,
+    R2Sql.of({
+      query: Effect.fn("R2Sql.query")(function* (query: string) {
+        const response = yield* Effect.tryPromise({
+          try: () =>
+            Bun.fetch(
+              `https://api.sql.cloudflarestorage.com/api/v1/accounts/${Resource.R2Sql.accountId}/r2-sql/query/${Resource.R2Sql.bucket}`,
+              {
+                method: "POST",
+                headers: {
+                  Authorization: `Bearer ${Resource.R2SqlAuthToken.value}`,
+                  "Content-Type": "application/json",
+                },
+                body: JSON.stringify({ query }),
+              },
+            ),
+          catch: (cause) => new R2SqlQueryError({ message: "Failed to run R2 SQL stats query", cause }),
+        })
+        const body = yield* Effect.tryPromise({
+          try: () => response.text(),
+          catch: (cause) =>
+            new R2SqlQueryError({ message: "Failed to read R2 SQL stats response", status: response.status, cause }),
+        })
+        const decoded = yield* decodeResponse(body).pipe(
+          Effect.mapError(
+            (cause) =>
+              new R2SqlQueryError({
+                message: "R2 SQL returned an invalid stats response",
+                status: response.status,
+                cause,
+              }),
+          ),
+        )
+        if (!response.ok || !decoded.success || !decoded.result)
+          return yield* Effect.fail(
+            new R2SqlQueryError({
+              message: `R2 SQL stats query failed: ${JSON.stringify(decoded.errors)}`,
+              requestId: decoded.result?.request_id,
+              status: response.status,
+            }),
+          )
+
+        // R2 SQL has no OFFSET support and caps LIMIT at 10,000. Each stats
+        // query is scoped to one day or week, and reaching the cap is treated as
+        // an error so a newly high-cardinality period can never be truncated.
+        if (decoded.result.rows.length >= R2_SQL_MAX_ROWS)
+          return yield* Effect.fail(
+            new R2SqlQueryError({
+              message: `R2 SQL stats query reached the ${R2_SQL_MAX_ROWS} row limit`,
+              requestId: decoded.result.request_id,
+              status: response.status,
+            }),
+          )
+
+        return decoded.result.rows.map((row) =>
+          Object.fromEntries(
+            Object.entries(row).flatMap(([key, value]) => (value === null ? [] : [[key, String(value)]])),
+          ),
+        )
+      }),
+    }),
+  )
+}
@@ -11,6 +11,17 @@ declare module "sst/resource" {
      type: "sst.sst.Linkable"
      workgroup: string
    }
+    R2Sql: {
+      accountId: string
+      bucket: string
+      namespace: string
+      table: string
+      type: "sst.sst.Linkable"
+    }
+    R2SqlAuthToken: {
+      type: "sst.sst.Secret"
+      value: string
+    }
    StatsSyncConfig: {
      dataset: string
      type: "sst.sst.Linkable"
@@ -1,12 +1,12 @@
 import { DateTime, Effect } from "effect"
 import { Resource } from "sst/resource"
-import { Athena, AthenaQueryError, AthenaQueryTimeoutError } from "./athena"
 import { DatabaseError } from "./database"
 import { GeoStatRepo, rowsFromAggregates as geoRowsFromAggregates } from "./domain/geo"
-import { buildStatsQuery, toGeoAggregate, toModelAggregate, toProviderAggregate } from "./domain/inference"
+import { buildStatsQueries, toGeoAggregate, toModelAggregate, toProviderAggregate } from "./domain/inference"
 import { ModelStatRepo, rowsFromAggregates as modelRowsFromAggregates } from "./domain/model"
 import { ProviderStatRepo, rowsFromAggregates as providerRowsFromAggregates } from "./domain/provider"
 import { startOfIsoWeek } from "./domain/stat"
+import { R2Sql, R2SqlQueryError } from "./r2-sql"

 const DATALAKE_INGESTION_LAG_MS = 5 * 60_000
 const STATS_DATA_START_MS = new Date("2026-05-28T00:00:00.000Z").getTime()
@@ -18,23 +18,25 @@ const DISPLAY_WINDOW_MS = 56 * 86_400_000
 const INCREMENTAL_LOOKBACK_MS = 2 * 3_600_000

 export type SyncStatsResult = { ok: true; rows: number; startedAt: string; periodStart: string; periodEnd: string }
-export type SyncStatsError = AthenaQueryError | AthenaQueryTimeoutError | DatabaseError
+export type SyncStatsError = R2SqlQueryError | DatabaseError

 export const syncStats: (options?: {
  full?: boolean
-}) => Effect.Effect<SyncStatsResult, SyncStatsError, Athena | ModelStatRepo | ProviderStatRepo | GeoStatRepo> =
+}) => Effect.Effect<SyncStatsResult, SyncStatsError, R2Sql | ModelStatRepo | ProviderStatRepo | GeoStatRepo> =
  Effect.fn("StatSync.sync")(function* (options?: { full?: boolean }) {
    const startedAt = yield* DateTime.nowAsDate
    const periodEnd = new Date(Math.floor((startedAt.getTime() - DATALAKE_INGESTION_LAG_MS) / 60_000) * 60_000)
    const periodStart = options?.full ? fullPeriodStart(periodEnd) : incrementalPeriodStart(periodEnd)
-    const athena = yield* Athena
+    const r2Sql = yield* R2Sql
    const modelStats = yield* ModelStatRepo
    const providerStats = yield* ProviderStatRepo
    const geoStats = yield* GeoStatRepo

    yield* logRuntimeCheck()

-    const rows = yield* athena.query(buildStatsQuery(periodStart, periodEnd))
+    const rows = yield* Effect.forEach(buildStatsQueries(periodStart, periodEnd), r2Sql.query, {
+      concurrency: 4,
+    }).pipe(Effect.map((batches) => batches.flat()))
    const modelRows = modelRowsFromAggregates(rows.filter((row) => row.dimension === "model").flatMap(toModelAggregate))
    const providerRows = providerRowsFromAggregates(
      rows.filter((row) => row.dimension === "provider").flatMap(toProviderAggregate),
@@ -77,7 +79,7 @@ export const syncStats: (options?: {
    }
  })

-// May 27 was partial, so keep Athena stats anchored at the first complete day.
+// May 27 was partial, so keep stats anchored at the first complete day.
 function fullPeriodStart(periodEnd: Date) {
  return new Date(
    Math.max(
@@ -99,13 +101,12 @@ function incrementalPeriodStart(periodEnd: Date) {

 function logRuntimeCheck() {
  return Effect.logInfo(
-    `athena stats runtime check ${JSON.stringify({
-      catalog: Resource.InferenceEvent.catalog,
-      database: Resource.InferenceEvent.database,
+    `r2 sql stats runtime check ${JSON.stringify({
+      accountId: Resource.R2Sql.accountId,
+      bucket: Resource.R2Sql.bucket,
      dataset: Resource.StatsSyncConfig.dataset,
-      table: Resource.InferenceEvent.table,
-      workgroup: Resource.InferenceEvent.workgroup,
-      region: Resource.InferenceEvent.region,
+      namespace: Resource.R2Sql.namespace,
+      table: Resource.R2Sql.table,
      stage: Resource.App.stage,
    })}`,
  )
@@ -1,7 +1,7 @@
 {
  "$schema": "https://json.schemastore.org/package.json",
  "name": "@opencode-ai/stats-server",
-  "version": "1.18.16",
+  "version": "1.18.17",
  "private": true,
  "type": "module",
  "license": "MIT",
@@ -1,6 +1,6 @@
 import * as NodeRuntime from "@effect/platform-node/NodeRuntime"
-import { Athena } from "@opencode-ai/stats-core/athena"
 import { ModelStatRepo } from "@opencode-ai/stats-core/domain/model"
+import { R2Sql } from "@opencode-ai/stats-core/r2-sql"
 import { layer as statsLayer } from "@opencode-ai/stats-core/runtime"
 import { syncStats } from "@opencode-ai/stats-core/stat-sync"
 import { Cause, Duration, Effect, Layer, Schedule } from "effect"
@@ -8,7 +8,7 @@ import { Cause, Duration, Effect, Layer, Schedule } from "effect"
 const SYNC_INTERVAL = "1 hour"
 const SYNC_INTERVAL_MS = 3_600_000

-const runtimeLayer = Layer.mergeAll(statsLayer, Athena.layer)
+const runtimeLayer = Layer.mergeAll(statsLayer, R2Sql.layer)

 const daemon = Effect.gen(function* () {
  yield* Effect.logInfo("stats sync daemon started")
@@ -40,9 +40,9 @@ const daemon = Effect.gen(function* () {
  yield* pass.pipe(Effect.repeat(Schedule.fixed(SYNC_INTERVAL)))
 }).pipe(Effect.forkScoped)

-// A restarted daemon must not immediately re-run the expensive Athena pass; resume
-// the hourly cadence from the last completed sync instead. This caps the Athena
-// spend of a crash loop at one pass per interval.
+// A restarted daemon must not immediately re-run the R2 SQL pass; resume the
+// hourly cadence from the last completed sync instead. This caps the query spend
+// of a crash loop at one pass per interval.
 const initialDelay = Effect.fnUntraced(function* () {
  const modelStats = yield* ModelStatRepo
  const lastSynced = yield* modelStats.lastSyncedAt().pipe(Effect.catchCause(() => Effect.succeed(null)))
@@ -1,7 +1,7 @@
 {
  "$schema": "https://json.schemastore.org/package.json",
  "name": "@opencode-ai/tui",
-  "version": "1.18.16",
+  "version": "1.18.17",
  "private": true,
  "type": "module",
  "license": "MIT",
@@ -1,6 +1,6 @@
 {
  "name": "@opencode-ai/ui",
-  "version": "1.18.16",
+  "version": "1.18.17",
  "type": "module",
  "license": "MIT",
  "repository": {
@@ -2,7 +2,7 @@
  "name": "@opencode-ai/web",
  "type": "module",
  "license": "MIT",
-  "version": "1.18.16",
+  "version": "1.18.17",
  "scripts": {
    "dev": "astro dev",
    "dev:remote": "VITE_API_URL=https://api.opencode.ai astro dev",
@@ -90,6 +90,7 @@ OpenCode Zen هي بوابة AI تتيح لك الوصول إلى هذه الن
 | Gemini 3.5 Flash Lite       | gemini-3.5-flash-lite       | `https://opencode.ai/zen/v1/models/gemini-3.5-flash-lite` | `@ai-sdk/google`            |
 | Gemini 3.1 Pro              | gemini-3.1-pro              | `https://opencode.ai/zen/v1/models/gemini-3.1-pro`        | `@ai-sdk/google`            |
 | Gemini 3 Flash              | gemini-3-flash              | `https://opencode.ai/zen/v1/models/gemini-3-flash`        | `@ai-sdk/google`            |
+| Grok 4.6                    | grok-4.6                    | `https://opencode.ai/zen/v1/responses`                    | `@ai-sdk/openai`            |
 | Grok 4.5                    | grok-4.5                    | `https://opencode.ai/zen/v1/responses`                    | `@ai-sdk/openai`            |
 | Grok Build 0.1              | grok-build-0.1              | `https://opencode.ai/zen/v1/responses`                    | `@ai-sdk/openai`            |
 | Qwen3.7 Max                 | qwen3.7-max                 | `https://opencode.ai/zen/v1/messages`                     | `@ai-sdk/anthropic`         |
@@ -178,6 +179,8 @@ https://opencode.ai/zen/v1/models
 | Gemini 3.1 Pro (≤ 200K tokens)    | $2.00   | $12.00  | $0.20           | -               |
 | Gemini 3.1 Pro (> 200K tokens)    | $4.00   | $18.00  | $0.40           | -               |
 | Gemini 3 Flash                    | $0.50   | $3.00   | $0.05           | -               |
+| Grok 4.6 (≤ 200K tokens)          | $2.00   | $6.00   | $0.50           | -               |
+| Grok 4.6 (> 200K tokens)          | $4.00   | $12.00  | $1.00           | -               |
 | Grok 4.5 (≤ 200K tokens)          | $2.00   | $6.00   | $0.30           | -               |
 | Grok 4.5 (> 200K tokens)          | $4.00   | $12.00  | $0.60           | -               |
 | Grok Build 0.1                    | $1.00   | $2.00   | $0.20           | -               |
@@ -95,6 +95,7 @@ Našim modelima možete pristupiti i preko sljedećih API endpointa.
 | Gemini 3.5 Flash Lite       | gemini-3.5-flash-lite       | `https://opencode.ai/zen/v1/models/gemini-3.5-flash-lite` | `@ai-sdk/google`            |
 | Gemini 3.1 Pro              | gemini-3.1-pro              | `https://opencode.ai/zen/v1/models/gemini-3.1-pro`        | `@ai-sdk/google`            |
 | Gemini 3 Flash              | gemini-3-flash              | `https://opencode.ai/zen/v1/models/gemini-3-flash`        | `@ai-sdk/google`            |
+| Grok 4.6                    | grok-4.6                    | `https://opencode.ai/zen/v1/responses`                    | `@ai-sdk/openai`            |
 | Grok 4.5                    | grok-4.5                    | `https://opencode.ai/zen/v1/responses`                    | `@ai-sdk/openai`            |
 | Grok Build 0.1              | grok-build-0.1              | `https://opencode.ai/zen/v1/responses`                    | `@ai-sdk/openai`            |
 | Qwen3.7 Max                 | qwen3.7-max                 | `https://opencode.ai/zen/v1/messages`                     | `@ai-sdk/anthropic`         |
@@ -185,6 +186,8 @@ Podržavamo pay-as-you-go model. Ispod su cijene **po 1M tokena**.
 | Gemini 3.1 Pro (≤ 200K tokens)    | $2.00  | $12.00  | $0.20       | -            |
 | Gemini 3.1 Pro (> 200K tokens)    | $4.00  | $18.00  | $0.40       | -            |
 | Gemini 3 Flash                    | $0.50  | $3.00   | $0.05       | -            |
+| Grok 4.6 (≤ 200K tokens)          | $2.00  | $6.00   | $0.50       | -            |
+| Grok 4.6 (> 200K tokens)          | $4.00  | $12.00  | $1.00       | -            |
 | Grok 4.5 (≤ 200K tokens)          | $2.00  | $6.00   | $0.30       | -            |
 | Grok 4.5 (> 200K tokens)          | $4.00  | $12.00  | $0.60       | -            |
 | Grok Build 0.1                    | $1.00  | $2.00   | $0.20       | -            |
@@ -95,6 +95,7 @@ Du kan også få adgang til vores modeller gennem følgende API-endpoints.
 | Gemini 3.5 Flash Lite       | gemini-3.5-flash-lite       | `https://opencode.ai/zen/v1/models/gemini-3.5-flash-lite` | `@ai-sdk/google`            |
 | Gemini 3.1 Pro              | gemini-3.1-pro              | `https://opencode.ai/zen/v1/models/gemini-3.1-pro`        | `@ai-sdk/google`            |
 | Gemini 3 Flash              | gemini-3-flash              | `https://opencode.ai/zen/v1/models/gemini-3-flash`        | `@ai-sdk/google`            |
+| Grok 4.6                    | grok-4.6                    | `https://opencode.ai/zen/v1/responses`                    | `@ai-sdk/openai`            |
 | Grok 4.5                    | grok-4.5                    | `https://opencode.ai/zen/v1/responses`                    | `@ai-sdk/openai`            |
 | Grok Build 0.1              | grok-build-0.1              | `https://opencode.ai/zen/v1/responses`                    | `@ai-sdk/openai`            |
 | Qwen3.7 Max                 | qwen3.7-max                 | `https://opencode.ai/zen/v1/messages`                     | `@ai-sdk/anthropic`         |
@@ -185,6 +186,8 @@ Vi understøtter en pay-as-you-go-model. Nedenfor er priserne **pr. 1M tokens**.
 | Gemini 3.1 Pro (≤ 200K tokens)    | $2.00  | $12.00  | $0.20       | -            |
 | Gemini 3.1 Pro (> 200K tokens)    | $4.00  | $18.00  | $0.40       | -            |
 | Gemini 3 Flash                    | $0.50  | $3.00   | $0.05       | -            |
+| Grok 4.6 (≤ 200K tokens)          | $2.00  | $6.00   | $0.50       | -            |
+| Grok 4.6 (> 200K tokens)          | $4.00  | $12.00  | $1.00       | -            |
 | Grok 4.5 (≤ 200K tokens)          | $2.00  | $6.00   | $0.30       | -            |
 | Grok 4.5 (> 200K tokens)          | $4.00  | $12.00  | $0.60       | -            |
 | Grok Build 0.1                    | $1.00  | $2.00   | $0.20       | -            |
@@ -86,6 +86,7 @@ Du kannst auch über die folgenden API-Endpunkte auf unsere Modelle zugreifen.
 | Gemini 3.5 Flash Lite       | gemini-3.5-flash-lite       | `https://opencode.ai/zen/v1/models/gemini-3.5-flash-lite` | `@ai-sdk/google`            |
 | Gemini 3.1 Pro              | gemini-3.1-pro              | `https://opencode.ai/zen/v1/models/gemini-3.1-pro`        | `@ai-sdk/google`            |
 | Gemini 3 Flash              | gemini-3-flash              | `https://opencode.ai/zen/v1/models/gemini-3-flash`        | `@ai-sdk/google`            |
+| Grok 4.6                    | grok-4.6                    | `https://opencode.ai/zen/v1/responses`                    | `@ai-sdk/openai`            |
 | Grok 4.5                    | grok-4.5                    | `https://opencode.ai/zen/v1/responses`                    | `@ai-sdk/openai`            |
 | Grok Build 0.1              | grok-build-0.1              | `https://opencode.ai/zen/v1/responses`                    | `@ai-sdk/openai`            |
 | Qwen3.7 Max                 | qwen3.7-max                 | `https://opencode.ai/zen/v1/messages`                     | `@ai-sdk/anthropic`         |
@@ -174,6 +175,8 @@ Wir unterstützen ein Pay-as-you-go-Modell. Unten findest du die Preise **pro 1M
 | Gemini 3.1 Pro (≤ 200K tokens)    | $2.00  | $12.00  | $0.20       | -            |
 | Gemini 3.1 Pro (> 200K tokens)    | $4.00  | $18.00  | $0.40       | -            |
 | Gemini 3 Flash                    | $0.50  | $3.00   | $0.05       | -            |
+| Grok 4.6 (≤ 200K tokens)          | $2.00  | $6.00   | $0.50       | -            |
+| Grok 4.6 (> 200K tokens)          | $4.00  | $12.00  | $1.00       | -            |
 | Grok 4.5 (≤ 200K tokens)          | $2.00  | $6.00   | $0.30       | -            |
 | Grok 4.5 (> 200K tokens)          | $4.00  | $12.00  | $0.60       | -            |
 | Grok Build 0.1                    | $1.00  | $2.00   | $0.20       | -            |
@@ -17,7 +17,7 @@ You can also check out [awesome-opencode](https://github.com/awesome-opencode/aw

 | Name                                                                                               | Description                                                                                        |
 | -------------------------------------------------------------------------------------------------- | -------------------------------------------------------------------------------------------------- |
-| [opencode-daytona](https://github.com/daytonaio/daytona/tree/main/libs/opencode-plugin)            | Automatically run OpenCode sessions in isolated Daytona sandboxes with git sync and live previews  |
+| [opencode-daytona](https://github.com/daytona/integrations/tree/main/packages/opencode-plugin)     | Automatically run OpenCode sessions in isolated Daytona sandboxes with git sync and live previews  |
 | [opencode-helicone-session](https://github.com/H2Shami/opencode-helicone-session)                  | Automatically inject Helicone session headers for request grouping                                 |
 | [opencode-type-inject](https://github.com/nick-vi/opencode-type-inject)                            | Auto-inject TypeScript/Svelte types into file reads with lookup tools                              |
 | [opencode-openai-codex-auth](https://github.com/numman-ali/opencode-openai-codex-auth)             | Use your ChatGPT Plus/Pro subscription instead of API credits                                      |
@@ -95,6 +95,7 @@ También puedes acceder a nuestros modelos a través de los siguientes endpoints
 | Gemini 3.5 Flash Lite       | gemini-3.5-flash-lite       | `https://opencode.ai/zen/v1/models/gemini-3.5-flash-lite` | `@ai-sdk/google`            |
 | Gemini 3.1 Pro              | gemini-3.1-pro              | `https://opencode.ai/zen/v1/models/gemini-3.1-pro`        | `@ai-sdk/google`            |
 | Gemini 3 Flash              | gemini-3-flash              | `https://opencode.ai/zen/v1/models/gemini-3-flash`        | `@ai-sdk/google`            |
+| Grok 4.6                    | grok-4.6                    | `https://opencode.ai/zen/v1/responses`                    | `@ai-sdk/openai`            |
 | Grok 4.5                    | grok-4.5                    | `https://opencode.ai/zen/v1/responses`                    | `@ai-sdk/openai`            |
 | Grok Build 0.1              | grok-build-0.1              | `https://opencode.ai/zen/v1/responses`                    | `@ai-sdk/openai`            |
 | Qwen3.7 Max                 | qwen3.7-max                 | `https://opencode.ai/zen/v1/messages`                     | `@ai-sdk/anthropic`         |
@@ -185,6 +186,8 @@ Admitimos un modelo de pago por uso. A continuación se muestran los precios **p
 | Gemini 3.1 Pro (≤ 200K tokens)    | $2.00   | $12.00  | $0.20            | -                  |
 | Gemini 3.1 Pro (> 200K tokens)    | $4.00   | $18.00  | $0.40            | -                  |
 | Gemini 3 Flash                    | $0.50   | $3.00   | $0.05            | -                  |
+| Grok 4.6 (≤ 200K tokens)          | $2.00   | $6.00   | $0.50            | -                  |
+| Grok 4.6 (> 200K tokens)          | $4.00   | $12.00  | $1.00            | -                  |
 | Grok 4.5 (≤ 200K tokens)          | $2.00   | $6.00   | $0.30            | -                  |
 | Grok 4.5 (> 200K tokens)          | $4.00   | $12.00  | $0.60            | -                  |
 | Grok Build 0.1                    | $1.00   | $2.00   | $0.20            | -                  |
@@ -86,6 +86,7 @@ Vous pouvez également accéder à nos modèles via les points de terminaison AP
 | Gemini 3.5 Flash Lite       | gemini-3.5-flash-lite       | `https://opencode.ai/zen/v1/models/gemini-3.5-flash-lite` | `@ai-sdk/google`            |
 | Gemini 3.1 Pro              | gemini-3.1-pro              | `https://opencode.ai/zen/v1/models/gemini-3.1-pro`        | `@ai-sdk/google`            |
 | Gemini 3 Flash              | gemini-3-flash              | `https://opencode.ai/zen/v1/models/gemini-3-flash`        | `@ai-sdk/google`            |
+| Grok 4.6                    | grok-4.6                    | `https://opencode.ai/zen/v1/responses`                    | `@ai-sdk/openai`            |
 | Grok 4.5                    | grok-4.5                    | `https://opencode.ai/zen/v1/responses`                    | `@ai-sdk/openai`            |
 | Grok Build 0.1              | grok-build-0.1              | `https://opencode.ai/zen/v1/responses`                    | `@ai-sdk/openai`            |
 | Qwen3.7 Max                 | qwen3.7-max                 | `https://opencode.ai/zen/v1/messages`                     | `@ai-sdk/anthropic`         |
@@ -174,6 +175,8 @@ Nous prenons en charge un modèle de paiement à l'utilisation. Vous trouverez c
 | Gemini 3.1 Pro (≤ 200K tokens)    | $2.00  | $12.00  | $0.20       | -            |
 | Gemini 3.1 Pro (> 200K tokens)    | $4.00  | $18.00  | $0.40       | -            |
 | Gemini 3 Flash                    | $0.50  | $3.00   | $0.05       | -            |
+| Grok 4.6 (≤ 200K tokens)          | $2.00  | $6.00   | $0.50       | -            |
+| Grok 4.6 (> 200K tokens)          | $4.00  | $12.00  | $1.00       | -            |
 | Grok 4.5 (≤ 200K tokens)          | $2.00  | $6.00   | $0.30       | -            |
 | Grok 4.5 (> 200K tokens)          | $4.00  | $12.00  | $0.60       | -            |
 | Grok Build 0.1                    | $1.00  | $2.00   | $0.20       | -            |
@@ -97,7 +97,7 @@ Or you can set it up manually.
    issues: write
  ```

-  You can also use a [personal access tokens](https://docs.github.com/en/authentication/keeping-your-account-and-data-secure/managing-your-personal-access-tokens)(PAT) if preferred.
+  You can also use a [personal access token](https://docs.github.com/en/authentication/keeping-your-account-and-data-secure/managing-your-personal-access-tokens)(PAT) if preferred.

 ---

@@ -252,13 +252,13 @@ https://opencode.ai/zen/go/v1/models
 | Qwen3.6 Plus      | Not used       | 0 days         |
 | MiniMax M3        | Not used       | 0 days         |
 | MiniMax M2.7      | Not used       | 0 days         |
-| DeepSeek V4 Pro   | Not used       | 0 days         |
-| DeepSeek V4 Flash | Not used       | 0 days         |
+| DeepSeek V4 Pro   | Not used       | 0 days\*       |
+| DeepSeek V4 Flash | Not used       | 0 days\*       |
 | Hy3               | Not used       | 0 days         |

 - **Grok 4.5:** ZDR disables important API features that depend on stored data, including the stateful Responses API, Files and Collections, and the Batch API. [Learn more](https://docs.x.ai/developers/faq/security#what-is-zero-data-retention-zdr).
 - **GPT 5.6 Luna:** Abuse monitoring logs are generated for all API feature usage and retained for up to 30 days. [Learn more](https://developers.openai.com/api/docs/guides/your-data#data-retention-controls-for-abuse-monitoring).
- **DeepSeek V4 Flash:** ZDR agreement is renewed monthly. The current agreement is valid through August 31, 2026.
+- **DeepSeek:** ZDR agreement is renewed monthly. The current agreement is valid through August 31, 2026.

 ---

@@ -95,6 +95,7 @@ Puoi anche accedere ai nostri modelli tramite i seguenti endpoint API.
 | Gemini 3.5 Flash Lite       | gemini-3.5-flash-lite       | `https://opencode.ai/zen/v1/models/gemini-3.5-flash-lite` | `@ai-sdk/google`            |
 | Gemini 3.1 Pro              | gemini-3.1-pro              | `https://opencode.ai/zen/v1/models/gemini-3.1-pro`        | `@ai-sdk/google`            |
 | Gemini 3 Flash              | gemini-3-flash              | `https://opencode.ai/zen/v1/models/gemini-3-flash`        | `@ai-sdk/google`            |
+| Grok 4.6                    | grok-4.6                    | `https://opencode.ai/zen/v1/responses`                    | `@ai-sdk/openai`            |
 | Grok 4.5                    | grok-4.5                    | `https://opencode.ai/zen/v1/responses`                    | `@ai-sdk/openai`            |
 | Grok Build 0.1              | grok-build-0.1              | `https://opencode.ai/zen/v1/responses`                    | `@ai-sdk/openai`            |
 | Qwen3.7 Max                 | qwen3.7-max                 | `https://opencode.ai/zen/v1/messages`                     | `@ai-sdk/anthropic`         |
@@ -185,6 +186,8 @@ Supportiamo un modello pay-as-you-go. Qui sotto trovi i prezzi **per 1M token**.
 | Gemini 3.1 Pro (≤ 200K tokens)    | $2.00  | $12.00  | $0.20       | -            |
 | Gemini 3.1 Pro (> 200K tokens)    | $4.00  | $18.00  | $0.40       | -            |
 | Gemini 3 Flash                    | $0.50  | $3.00   | $0.05       | -            |
+| Grok 4.6 (≤ 200K tokens)          | $2.00  | $6.00   | $0.50       | -            |
+| Grok 4.6 (> 200K tokens)          | $4.00  | $12.00  | $1.00       | -            |
 | Grok 4.5 (≤ 200K tokens)          | $2.00  | $6.00   | $0.30       | -            |
 | Grok 4.5 (> 200K tokens)          | $4.00  | $12.00  | $0.60       | -            |
 | Grok Build 0.1                    | $1.00  | $2.00   | $0.20       | -            |
@@ -86,6 +86,7 @@ OpenCode Zen は、OpenCode のほかのプロバイダーと同じように動
 | Gemini 3.5 Flash Lite       | gemini-3.5-flash-lite       | `https://opencode.ai/zen/v1/models/gemini-3.5-flash-lite` | `@ai-sdk/google`            |
 | Gemini 3.1 Pro              | gemini-3.1-pro              | `https://opencode.ai/zen/v1/models/gemini-3.1-pro`        | `@ai-sdk/google`            |
 | Gemini 3 Flash              | gemini-3-flash              | `https://opencode.ai/zen/v1/models/gemini-3-flash`        | `@ai-sdk/google`            |
+| Grok 4.6                    | grok-4.6                    | `https://opencode.ai/zen/v1/responses`                    | `@ai-sdk/openai`            |
 | Grok 4.5                    | grok-4.5                    | `https://opencode.ai/zen/v1/responses`                    | `@ai-sdk/openai`            |
 | Grok Build 0.1              | grok-build-0.1              | `https://opencode.ai/zen/v1/responses`                    | `@ai-sdk/openai`            |
 | Qwen3.7 Max                 | qwen3.7-max                 | `https://opencode.ai/zen/v1/messages`                     | `@ai-sdk/anthropic`         |
@@ -174,6 +175,8 @@ https://opencode.ai/zen/v1/models
 | Gemini 3.1 Pro (≤ 200K tokens)    | $2.00  | $12.00  | $0.20       | -            |
 | Gemini 3.1 Pro (> 200K tokens)    | $4.00  | $18.00  | $0.40       | -            |
 | Gemini 3 Flash                    | $0.50  | $3.00   | $0.05       | -            |
+| Grok 4.6 (≤ 200K tokens)          | $2.00  | $6.00   | $0.50       | -            |
+| Grok 4.6 (> 200K tokens)          | $4.00  | $12.00  | $1.00       | -            |
 | Grok 4.5 (≤ 200K tokens)          | $2.00  | $6.00   | $0.30       | -            |
 | Grok 4.5 (> 200K tokens)          | $4.00  | $12.00  | $0.60       | -            |
 | Grok Build 0.1                    | $1.00  | $2.00   | $0.20       | -            |
@@ -86,6 +86,7 @@ OpenCode Zen은 OpenCode의 다른 provider와 똑같이 작동합니다.
 | Gemini 3.5 Flash Lite       | gemini-3.5-flash-lite       | `https://opencode.ai/zen/v1/models/gemini-3.5-flash-lite` | `@ai-sdk/google`            |
 | Gemini 3.1 Pro              | gemini-3.1-pro              | `https://opencode.ai/zen/v1/models/gemini-3.1-pro`        | `@ai-sdk/google`            |
 | Gemini 3 Flash              | gemini-3-flash              | `https://opencode.ai/zen/v1/models/gemini-3-flash`        | `@ai-sdk/google`            |
+| Grok 4.6                    | grok-4.6                    | `https://opencode.ai/zen/v1/responses`                    | `@ai-sdk/openai`            |
 | Grok 4.5                    | grok-4.5                    | `https://opencode.ai/zen/v1/responses`                    | `@ai-sdk/openai`            |
 | Grok Build 0.1              | grok-build-0.1              | `https://opencode.ai/zen/v1/responses`                    | `@ai-sdk/openai`            |
 | Qwen3.7 Max                 | qwen3.7-max                 | `https://opencode.ai/zen/v1/messages`                     | `@ai-sdk/anthropic`         |
@@ -174,6 +175,8 @@ https://opencode.ai/zen/v1/models
 | Gemini 3.1 Pro (≤ 200K tokens)    | $2.00  | $12.00  | $0.20       | -            |
 | Gemini 3.1 Pro (> 200K tokens)    | $4.00  | $18.00  | $0.40       | -            |
 | Gemini 3 Flash                    | $0.50  | $3.00   | $0.05       | -            |
+| Grok 4.6 (≤ 200K tokens)          | $2.00  | $6.00   | $0.50       | -            |
+| Grok 4.6 (> 200K tokens)          | $4.00  | $12.00  | $1.00       | -            |
 | Grok 4.5 (≤ 200K tokens)          | $2.00  | $6.00   | $0.30       | -            |
 | Grok 4.5 (> 200K tokens)          | $4.00  | $12.00  | $0.60       | -            |
 | Grok Build 0.1                    | $1.00  | $2.00   | $0.20       | -            |
@@ -95,6 +95,7 @@ Du kan også få tilgang til modellene våre gjennom følgende API-endepunkter.
 | Gemini 3.5 Flash Lite       | gemini-3.5-flash-lite       | `https://opencode.ai/zen/v1/models/gemini-3.5-flash-lite` | `@ai-sdk/google`            |
 | Gemini 3.1 Pro              | gemini-3.1-pro              | `https://opencode.ai/zen/v1/models/gemini-3.1-pro`        | `@ai-sdk/google`            |
 | Gemini 3 Flash              | gemini-3-flash              | `https://opencode.ai/zen/v1/models/gemini-3-flash`        | `@ai-sdk/google`            |
+| Grok 4.6                    | grok-4.6                    | `https://opencode.ai/zen/v1/responses`                    | `@ai-sdk/openai`            |
 | Grok 4.5                    | grok-4.5                    | `https://opencode.ai/zen/v1/responses`                    | `@ai-sdk/openai`            |
 | Grok Build 0.1              | grok-build-0.1              | `https://opencode.ai/zen/v1/responses`                    | `@ai-sdk/openai`            |
 | Qwen3.7 Max                 | qwen3.7-max                 | `https://opencode.ai/zen/v1/messages`                     | `@ai-sdk/anthropic`         |
@@ -185,6 +186,8 @@ Vi støtter en pay-as-you-go-modell. Nedenfor er prisene **per 1M tokens**.
 | Gemini 3.1 Pro (≤ 200K tokens)    | $2.00   | $12.00  | $0.20         | -               |
 | Gemini 3.1 Pro (> 200K tokens)    | $4.00   | $18.00  | $0.40         | -               |
 | Gemini 3 Flash                    | $0.50   | $3.00   | $0.05         | -               |
+| Grok 4.6 (≤ 200K tokens)          | $2.00   | $6.00   | $0.50         | -               |
+| Grok 4.6 (> 200K tokens)          | $4.00   | $12.00  | $1.00         | -               |
 | Grok 4.5 (≤ 200K tokens)          | $2.00   | $6.00   | $0.30         | -               |
 | Grok 4.5 (> 200K tokens)          | $4.00   | $12.00  | $0.60         | -               |
 | Grok Build 0.1                    | $1.00   | $2.00   | $0.20         | -               |
@@ -95,6 +95,7 @@ Możesz też uzyskać dostęp do naszych modeli przez poniższe endpointy API.
 | Gemini 3.5 Flash Lite       | gemini-3.5-flash-lite       | `https://opencode.ai/zen/v1/models/gemini-3.5-flash-lite` | `@ai-sdk/google`            |
 | Gemini 3.1 Pro              | gemini-3.1-pro              | `https://opencode.ai/zen/v1/models/gemini-3.1-pro`        | `@ai-sdk/google`            |
 | Gemini 3 Flash              | gemini-3-flash              | `https://opencode.ai/zen/v1/models/gemini-3-flash`        | `@ai-sdk/google`            |
+| Grok 4.6                    | grok-4.6                    | `https://opencode.ai/zen/v1/responses`                    | `@ai-sdk/openai`            |
 | Grok 4.5                    | grok-4.5                    | `https://opencode.ai/zen/v1/responses`                    | `@ai-sdk/openai`            |
 | Grok Build 0.1              | grok-build-0.1              | `https://opencode.ai/zen/v1/responses`                    | `@ai-sdk/openai`            |
 | Qwen3.7 Max                 | qwen3.7-max                 | `https://opencode.ai/zen/v1/messages`                     | `@ai-sdk/anthropic`         |
@@ -185,6 +186,8 @@ Obsługujemy model pay-as-you-go. Poniżej znajdują się ceny **za 1M tokenów*
 | Gemini 3.1 Pro (≤ 200K tokens)    | $2.00   | $12.00  | $0.20          | -              |
 | Gemini 3.1 Pro (> 200K tokens)    | $4.00   | $18.00  | $0.40          | -              |
 | Gemini 3 Flash                    | $0.50   | $3.00   | $0.05          | -              |
+| Grok 4.6 (≤ 200K tokens)          | $2.00   | $6.00   | $0.50          | -              |
+| Grok 4.6 (> 200K tokens)          | $4.00   | $12.00  | $1.00          | -              |
 | Grok 4.5 (≤ 200K tokens)          | $2.00   | $6.00   | $0.30          | -              |
 | Grok 4.5 (> 200K tokens)          | $4.00   | $12.00  | $0.60          | -              |
 | Grok Build 0.1                    | $1.00   | $2.00   | $0.20          | -              |
@@ -759,7 +759,7 @@ Cloudflare Workers AI lets you run AI models on Cloudflare's global network dire

 ### DigitalOcean

-DigitalOcean's [Inference Engine](https://docs.digitalocean.com/products/inference/) provides access to open models like GPT-OSS, Llama, Qwen, and DeepSeek, plus custom [Inference Routers](https://docs.digitalocean.com/products/genai-platform/concepts/inference-routers/) that route each request to the cheapest, fastest, or best-fit model for a task.
+DigitalOcean's [Inference Engine](https://docs.digitalocean.com/products/inference/) provides access to open models like GPT-OSS, Llama, Qwen, and DeepSeek, plus custom [Inference Routers](https://docs.digitalocean.com/products/inference/how-to/use-inference-router/) that route each request to the cheapest, fastest, or best-fit model for a task.

 OpenCode supports two authentication methods:

@@ -2487,7 +2487,7 @@ You can use any OpenAI-compatible provider with opencode. Most modern AI provide
     "provider": {
       "myprovider": {
         "npm": "@ai-sdk/openai-compatible",
-         "name": "My AI ProviderDisplay Name",
+         "name": "My AI Provider Display Name",
         "options": {
           "baseURL": "https://api.myprovider.com/v1"
         },
@@ -2525,7 +2525,7 @@ Here's an example setting the `apiKey`, `headers`, and model `limit` options.
  "provider": {
    "myprovider": {
      "npm": "@ai-sdk/openai-compatible",
-      "name": "My AI ProviderDisplay Name",
+      "name": "My AI Provider Display Name",
      "options": {
        "baseURL": "https://api.myprovider.com/v1",
        "apiKey": "{env:ANTHROPIC_API_KEY}",
@@ -86,6 +86,7 @@ Você também pode acessar nossos modelos pelos seguintes endpoints de API.
 | Gemini 3.5 Flash Lite       | gemini-3.5-flash-lite       | `https://opencode.ai/zen/v1/models/gemini-3.5-flash-lite` | `@ai-sdk/google`            |
 | Gemini 3.1 Pro              | gemini-3.1-pro              | `https://opencode.ai/zen/v1/models/gemini-3.1-pro`        | `@ai-sdk/google`            |
 | Gemini 3 Flash              | gemini-3-flash              | `https://opencode.ai/zen/v1/models/gemini-3-flash`        | `@ai-sdk/google`            |
+| Grok 4.6                    | grok-4.6                    | `https://opencode.ai/zen/v1/responses`                    | `@ai-sdk/openai`            |
 | Grok 4.5                    | grok-4.5                    | `https://opencode.ai/zen/v1/responses`                    | `@ai-sdk/openai`            |
 | Grok Build 0.1              | grok-build-0.1              | `https://opencode.ai/zen/v1/responses`                    | `@ai-sdk/openai`            |
 | Qwen3.7 Max                 | qwen3.7-max                 | `https://opencode.ai/zen/v1/messages`                     | `@ai-sdk/anthropic`         |
@@ -174,6 +175,8 @@ Oferecemos um modelo pay-as-you-go. Abaixo estão os preços **por 1M tokens**.
 | Gemini 3.1 Pro (≤ 200K tokens)    | $2.00   | $12.00  | $0.20            | -                |
 | Gemini 3.1 Pro (> 200K tokens)    | $4.00   | $18.00  | $0.40            | -                |
 | Gemini 3 Flash                    | $0.50   | $3.00   | $0.05            | -                |
+| Grok 4.6 (≤ 200K tokens)          | $2.00   | $6.00   | $0.50            | -                |
+| Grok 4.6 (> 200K tokens)          | $4.00   | $12.00  | $1.00            | -                |
 | Grok 4.5 (≤ 200K tokens)          | $2.00   | $6.00   | $0.30            | -                |
 | Grok 4.5 (> 200K tokens)          | $4.00   | $12.00  | $0.60            | -                |
 | Grok Build 0.1                    | $1.00   | $2.00   | $0.20            | -                |
@@ -95,6 +95,7 @@ OpenCode Zen работает как любой другой провайдер
 | Gemini 3.5 Flash Lite       | gemini-3.5-flash-lite       | `https://opencode.ai/zen/v1/models/gemini-3.5-flash-lite` | `@ai-sdk/google`            |
 | Gemini 3.1 Pro              | gemini-3.1-pro              | `https://opencode.ai/zen/v1/models/gemini-3.1-pro`        | `@ai-sdk/google`            |
 | Gemini 3 Flash              | gemini-3-flash              | `https://opencode.ai/zen/v1/models/gemini-3-flash`        | `@ai-sdk/google`            |
+| Grok 4.6                    | grok-4.6                    | `https://opencode.ai/zen/v1/responses`                    | `@ai-sdk/openai`            |
 | Grok 4.5                    | grok-4.5                    | `https://opencode.ai/zen/v1/responses`                    | `@ai-sdk/openai`            |
 | Grok Build 0.1              | grok-build-0.1              | `https://opencode.ai/zen/v1/responses`                    | `@ai-sdk/openai`            |
 | Qwen3.7 Max                 | qwen3.7-max                 | `https://opencode.ai/zen/v1/messages`                     | `@ai-sdk/anthropic`         |
@@ -185,6 +186,8 @@ https://opencode.ai/zen/v1/models
 | Gemini 3.1 Pro (≤ 200K tokens)    | $2.00  | $12.00  | $0.20       | -            |
 | Gemini 3.1 Pro (> 200K tokens)    | $4.00  | $18.00  | $0.40       | -            |
 | Gemini 3 Flash                    | $0.50  | $3.00   | $0.05       | -            |
+| Grok 4.6 (≤ 200K tokens)          | $2.00  | $6.00   | $0.50       | -            |
+| Grok 4.6 (> 200K tokens)          | $4.00  | $12.00  | $1.00       | -            |
 | Grok 4.5 (≤ 200K tokens)          | $2.00  | $6.00   | $0.30       | -            |
 | Grok 4.5 (> 200K tokens)          | $4.00  | $12.00  | $0.60       | -            |
 | Grok Build 0.1                    | $1.00  | $2.00   | $0.20       | -            |
@@ -88,6 +88,7 @@ OpenCode Zen ทำงานเหมือน provider อื่น ๆ ใน
 | Gemini 3.5 Flash Lite       | gemini-3.5-flash-lite       | `https://opencode.ai/zen/v1/models/gemini-3.5-flash-lite` | `@ai-sdk/google`            |
 | Gemini 3.1 Pro              | gemini-3.1-pro              | `https://opencode.ai/zen/v1/models/gemini-3.1-pro`        | `@ai-sdk/google`            |
 | Gemini 3 Flash              | gemini-3-flash              | `https://opencode.ai/zen/v1/models/gemini-3-flash`        | `@ai-sdk/google`            |
+| Grok 4.6                    | grok-4.6                    | `https://opencode.ai/zen/v1/responses`                    | `@ai-sdk/openai`            |
 | Grok 4.5                    | grok-4.5                    | `https://opencode.ai/zen/v1/responses`                    | `@ai-sdk/openai`            |
 | Grok Build 0.1              | grok-build-0.1              | `https://opencode.ai/zen/v1/responses`                    | `@ai-sdk/openai`            |
 | Qwen3.7 Max                 | qwen3.7-max                 | `https://opencode.ai/zen/v1/messages`                     | `@ai-sdk/anthropic`         |
@@ -176,6 +177,8 @@ https://opencode.ai/zen/v1/models
 | Gemini 3.1 Pro (≤ 200K tokens)    | $2.00  | $12.00  | $0.20       | -            |
 | Gemini 3.1 Pro (> 200K tokens)    | $4.00  | $18.00  | $0.40       | -            |
 | Gemini 3 Flash                    | $0.50  | $3.00   | $0.05       | -            |
+| Grok 4.6 (≤ 200K tokens)          | $2.00  | $6.00   | $0.50       | -            |
+| Grok 4.6 (> 200K tokens)          | $4.00  | $12.00  | $1.00       | -            |
 | Grok 4.5 (≤ 200K tokens)          | $2.00  | $6.00   | $0.30       | -            |
 | Grok 4.5 (> 200K tokens)          | $4.00  | $12.00  | $0.60       | -            |
 | Grok Build 0.1                    | $1.00  | $2.00   | $0.20       | -            |
@@ -86,6 +86,7 @@ Modellerimize aşağıdaki API uç noktaları aracılığıyla da erişebilirsin
 | Gemini 3.5 Flash Lite       | gemini-3.5-flash-lite       | `https://opencode.ai/zen/v1/models/gemini-3.5-flash-lite` | `@ai-sdk/google`            |
 | Gemini 3.1 Pro              | gemini-3.1-pro              | `https://opencode.ai/zen/v1/models/gemini-3.1-pro`        | `@ai-sdk/google`            |
 | Gemini 3 Flash              | gemini-3-flash              | `https://opencode.ai/zen/v1/models/gemini-3-flash`        | `@ai-sdk/google`            |
+| Grok 4.6                    | grok-4.6                    | `https://opencode.ai/zen/v1/responses`                    | `@ai-sdk/openai`            |
 | Grok 4.5                    | grok-4.5                    | `https://opencode.ai/zen/v1/responses`                    | `@ai-sdk/openai`            |
 | Grok Build 0.1              | grok-build-0.1              | `https://opencode.ai/zen/v1/responses`                    | `@ai-sdk/openai`            |
 | Qwen3.7 Max                 | qwen3.7-max                 | `https://opencode.ai/zen/v1/messages`                     | `@ai-sdk/anthropic`         |
@@ -174,6 +175,8 @@ Kullandıkça öde modelini destekliyoruz. Aşağıda **1M token başına** fiya
 | Gemini 3.1 Pro (≤ 200K tokens)    | $2.00  | $12.00  | $0.20       | -            |
 | Gemini 3.1 Pro (> 200K tokens)    | $4.00  | $18.00  | $0.40       | -            |
 | Gemini 3 Flash                    | $0.50  | $3.00   | $0.05       | -            |
+| Grok 4.6 (≤ 200K tokens)          | $2.00  | $6.00   | $0.50       | -            |
+| Grok 4.6 (> 200K tokens)          | $4.00  | $12.00  | $1.00       | -            |
 | Grok 4.5 (≤ 200K tokens)          | $2.00  | $6.00   | $0.30       | -            |
 | Grok 4.5 (> 200K tokens)          | $4.00  | $12.00  | $0.60       | -            |
 | Grok Build 0.1                    | $1.00  | $2.00   | $0.20       | -            |
@@ -95,6 +95,7 @@ You can also access our models through the following API endpoints.
 | Gemini 3.5 Flash Lite       | gemini-3.5-flash-lite       | `https://opencode.ai/zen/v1/models/gemini-3.5-flash-lite` | `@ai-sdk/google`            |
 | Gemini 3.1 Pro              | gemini-3.1-pro              | `https://opencode.ai/zen/v1/models/gemini-3.1-pro`        | `@ai-sdk/google`            |
 | Gemini 3 Flash              | gemini-3-flash              | `https://opencode.ai/zen/v1/models/gemini-3-flash`        | `@ai-sdk/google`            |
+| Grok 4.6                    | grok-4.6                    | `https://opencode.ai/zen/v1/responses`                    | `@ai-sdk/openai`            |
 | Grok 4.5                    | grok-4.5                    | `https://opencode.ai/zen/v1/responses`                    | `@ai-sdk/openai`            |
 | Grok Build 0.1              | grok-build-0.1              | `https://opencode.ai/zen/v1/responses`                    | `@ai-sdk/openai`            |
 | Qwen3.7 Max                 | qwen3.7-max                 | `https://opencode.ai/zen/v1/messages`                     | `@ai-sdk/anthropic`         |
@@ -185,6 +186,8 @@ We support a pay-as-you-go model. Below are the prices **per 1M tokens**.
 | Gemini 3.1 Pro (≤ 200K tokens)    | $2.00  | $12.00  | $0.20       | -            |
 | Gemini 3.1 Pro (> 200K tokens)    | $4.00  | $18.00  | $0.40       | -            |
 | Gemini 3 Flash                    | $0.50  | $3.00   | $0.05       | -            |
+| Grok 4.6 (≤ 200K tokens)          | $2.00  | $6.00   | $0.50       | -            |
+| Grok 4.6 (> 200K tokens)          | $4.00  | $12.00  | $1.00       | -            |
 | Grok 4.5 (≤ 200K tokens)          | $2.00  | $6.00   | $0.30       | -            |
 | Grok 4.5 (> 200K tokens)          | $4.00  | $12.00  | $0.60       | -            |
 | Grok Build 0.1                    | $1.00  | $2.00   | $0.20       | -            |
@@ -86,6 +86,7 @@ OpenCode Zen 的工作方式与 OpenCode 中的任何其他提供商相同。
 | Gemini 3.5 Flash Lite       | gemini-3.5-flash-lite       | `https://opencode.ai/zen/v1/models/gemini-3.5-flash-lite` | `@ai-sdk/google`            |
 | Gemini 3.1 Pro              | gemini-3.1-pro              | `https://opencode.ai/zen/v1/models/gemini-3.1-pro`        | `@ai-sdk/google`            |
 | Gemini 3 Flash              | gemini-3-flash              | `https://opencode.ai/zen/v1/models/gemini-3-flash`        | `@ai-sdk/google`            |
+| Grok 4.6                    | grok-4.6                    | `https://opencode.ai/zen/v1/responses`                    | `@ai-sdk/openai`            |
 | Grok 4.5                    | grok-4.5                    | `https://opencode.ai/zen/v1/responses`                    | `@ai-sdk/openai`            |
 | Grok Build 0.1              | grok-build-0.1              | `https://opencode.ai/zen/v1/responses`                    | `@ai-sdk/openai`            |
 | Qwen3.7 Max                 | qwen3.7-max                 | `https://opencode.ai/zen/v1/messages`                     | `@ai-sdk/anthropic`         |
@@ -174,6 +175,8 @@ https://opencode.ai/zen/v1/models
 | Gemini 3.1 Pro (≤ 200K tokens)    | $2.00  | $12.00  | $0.20    | -        |
 | Gemini 3.1 Pro (> 200K tokens)    | $4.00  | $18.00  | $0.40    | -        |
 | Gemini 3 Flash                    | $0.50  | $3.00   | $0.05    | -        |
+| Grok 4.6 (≤ 200K tokens)          | $2.00  | $6.00   | $0.50    | -        |
+| Grok 4.6 (> 200K tokens)          | $4.00  | $12.00  | $1.00    | -        |
 | Grok 4.5 (≤ 200K tokens)          | $2.00  | $6.00   | $0.30    | -        |
 | Grok 4.5 (> 200K tokens)          | $4.00  | $12.00  | $0.60    | -        |
 | Grok Build 0.1                    | $1.00  | $2.00   | $0.20    | -        |
@@ -90,6 +90,7 @@ OpenCode Zen 的運作方式和 OpenCode 中的其他供應商一樣。
 | Gemini 3.5 Flash Lite       | gemini-3.5-flash-lite       | `https://opencode.ai/zen/v1/models/gemini-3.5-flash-lite` | `@ai-sdk/google`            |
 | Gemini 3.1 Pro              | gemini-3.1-pro              | `https://opencode.ai/zen/v1/models/gemini-3.1-pro`        | `@ai-sdk/google`            |
 | Gemini 3 Flash              | gemini-3-flash              | `https://opencode.ai/zen/v1/models/gemini-3-flash`        | `@ai-sdk/google`            |
+| Grok 4.6                    | grok-4.6                    | `https://opencode.ai/zen/v1/responses`                    | `@ai-sdk/openai`            |
 | Grok 4.5                    | grok-4.5                    | `https://opencode.ai/zen/v1/responses`                    | `@ai-sdk/openai`            |
 | Grok Build 0.1              | grok-build-0.1              | `https://opencode.ai/zen/v1/responses`                    | `@ai-sdk/openai`            |
 | Qwen3.7 Max                 | qwen3.7-max                 | `https://opencode.ai/zen/v1/messages`                     | `@ai-sdk/anthropic`         |
@@ -179,6 +180,8 @@ https://opencode.ai/zen/v1/models
 | Gemini 3.1 Pro (≤ 200K tokens)    | $2.00  | $12.00  | $0.20    | -        |
 | Gemini 3.1 Pro (> 200K tokens)    | $4.00  | $18.00  | $0.40    | -        |
 | Gemini 3 Flash                    | $0.50  | $3.00   | $0.05    | -        |
+| Grok 4.6 (≤ 200K tokens)          | $2.00  | $6.00   | $0.50    | -        |
+| Grok 4.6 (> 200K tokens)          | $4.00  | $12.00  | $1.00    | -        |
 | Grok 4.5 (≤ 200K tokens)          | $2.00  | $6.00   | $0.30    | -        |
 | Grok 4.5 (> 200K tokens)          | $4.00  | $12.00  | $0.60    | -        |
 | Grok Build 0.1                    | $1.00  | $2.00   | $0.20    | -        |
@@ -0,0 +1,79 @@
+diff --git a/dist/index.d.mts b/dist/index.d.mts
+index 8b23996dcce6c1ad5b17ef59f92196fb97312d79..80be2e52a347042b89da8e502834afd92120877a 100644
+--- a/dist/index.d.mts
+++ b/dist/index.d.mts
+@@ -10,13 +10,7 @@ declare const groqLanguageModelOptions: z.ZodObject<{
+         raw: "raw";
+         hidden: "hidden";
+     }>>;
+-    reasoningEffort: z.ZodOptional<z.ZodEnum<{
+-        none: "none";
+-        default: "default";
+-        low: "low";
+-        medium: "medium";
+-        high: "high";
+-    }>>;
+    reasoningEffort: z.ZodOptional<z.ZodString>;
+     parallelToolCalls: z.ZodOptional<z.ZodBoolean>;
+     user: z.ZodOptional<z.ZodString>;
+     structuredOutputs: z.ZodOptional<z.ZodBoolean>;
+diff --git a/dist/index.d.ts b/dist/index.d.ts
+index 8b23996dcce6c1ad5b17ef59f92196fb97312d79..80be2e52a347042b89da8e502834afd92120877a 100644
+--- a/dist/index.d.ts
+++ b/dist/index.d.ts
+@@ -10,13 +10,7 @@ declare const groqLanguageModelOptions: z.ZodObject<{
+         raw: "raw";
+         hidden: "hidden";
+     }>>;
+-    reasoningEffort: z.ZodOptional<z.ZodEnum<{
+-        none: "none";
+-        default: "default";
+-        low: "low";
+-        medium: "medium";
+-        high: "high";
+-    }>>;
+    reasoningEffort: z.ZodOptional<z.ZodString>;
+     parallelToolCalls: z.ZodOptional<z.ZodBoolean>;
+     user: z.ZodOptional<z.ZodString>;
+     structuredOutputs: z.ZodOptional<z.ZodBoolean>;
+diff --git a/dist/index.js b/dist/index.js
+index 45a104f2e0775761858eac2a82ced64bceba1f5e..f60ac36f4a064d527e8f8881b1d6c58ff69286a3 100644
+--- a/dist/index.js
+++ b/dist/index.js
+@@ -214,7 +214,7 @@ var groqLanguageModelOptions = import_v4.z.object({
+    * Specifies the reasoning effort level for model inference.
+    * @see https://console.groq.com/docs/reasoning#reasoning-effort
+    */
+-  reasoningEffort: import_v4.z.enum(["none", "default", "low", "medium", "high"]).optional(),
+  reasoningEffort: import_v4.z.string().optional(),
+   /**
+    * Whether to enable parallel function calling during tool use. Default to true.
+    */
+diff --git a/dist/index.mjs b/dist/index.mjs
+index c644c32235d8fa88c51c0fc6958feb1da4877c96..2c2f81869673eb4633e843d93ff5376abf1e67d0 100644
+--- a/dist/index.mjs
+++ b/dist/index.mjs
+@@ -203,7 +203,7 @@ var groqLanguageModelOptions = z.object({
+    * Specifies the reasoning effort level for model inference.
+    * @see https://console.groq.com/docs/reasoning#reasoning-effort
+    */
+-  reasoningEffort: z.enum(["none", "default", "low", "medium", "high"]).optional(),
+  reasoningEffort: z.string().optional(),
+   /**
+    * Whether to enable parallel function calling during tool use. Default to true.
+    */
+diff --git a/src/groq-chat-options.ts b/src/groq-chat-options.ts
+index 3812cdf53308709f166f05c58c5d46a5d8189c8b..af520c5459bd752b3cce03c3b4afbeed31157d90 100644
+--- a/src/groq-chat-options.ts
+++ b/src/groq-chat-options.ts
+@@ -33,9 +33,7 @@ export const groqLanguageModelOptions = z.object({
+    * Specifies the reasoning effort level for model inference.
+    * @see https://console.groq.com/docs/reasoning#reasoning-effort
+    */
+-  reasoningEffort: z
+-    .enum(['none', 'default', 'low', 'medium', 'high'])
+-    .optional(),
+  reasoningEffort: z.string().optional(),
+ 
+   /**
+    * Whether to enable parallel function calling during tool use. Default to true.
@@ -2,10 +2,12 @@ diff --git a/dist/index.d.mts b/dist/index.d.mts
 index 1bde0b9f8cbe6771a52c1041095c9dddfe8e5b6c..0ca2ffb2a0c9327aed5ddcf0004500dc8b42569f 100644
 --- a/dist/index.d.mts
 +++ b/dist/index.d.mts
-@@ -14,6 +14,7 @@ declare const mistralLanguageModelOptions: z.ZodObject<{
-         none: "none";
-         high: "high";
-     }>>;
+@@ -13,7 +13,5 @@ declare const mistralLanguageModelOptions: z.ZodObject<{
+-    reasoningEffort: z.ZodOptional<z.ZodEnum<{
+-        none: "none";
+-        high: "high";
+-    }>>;
+    reasoningEffort: z.ZodOptional<z.ZodString>;
 +    promptCacheKey: z.ZodOptional<z.ZodString>;
 }, z.core.$strip>;
 type MistralLanguageModelOptions = z.infer<typeof mistralLanguageModelOptions>;
@@ -14,10 +16,12 @@ diff --git a/dist/index.d.ts b/dist/index.d.ts
 index 1bde0b9f8cbe6771a52c1041095c9dddfe8e5b6c..0ca2ffb2a0c9327aed5ddcf0004500dc8b42569f 100644
 --- a/dist/index.d.ts
 +++ b/dist/index.d.ts
-@@ -14,6 +14,7 @@ declare const mistralLanguageModelOptions: z.ZodObject<{
-         none: "none";
-         high: "high";
-     }>>;
+@@ -13,7 +13,5 @@ declare const mistralLanguageModelOptions: z.ZodObject<{
+-    reasoningEffort: z.ZodOptional<z.ZodEnum<{
+-        none: "none";
+-        high: "high";
+-    }>>;
+    reasoningEffort: z.ZodOptional<z.ZodString>;
 +    promptCacheKey: z.ZodOptional<z.ZodString>;
 }, z.core.$strip>;
 type MistralLanguageModelOptions = z.infer<typeof mistralLanguageModelOptions>;
@@ -69,7 +73,7 @@ index d3f904c12a1d582cc7b9e9a2d30273e1a8505b28..267f34e20ea392b7a85ad5259d72d506
    * - `'none'`: Disable reasoning
    */
 -  reasoningEffort: import_v4.z.enum(["high", "none"]).optional()
-+  reasoningEffort: import_v4.z.enum(["high", "none"]).optional(),
+  reasoningEffort: import_v4.z.string().optional(),
 +  promptCacheKey: import_v4.z.string().optional()
 });
 
@@ -268,7 +272,7 @@ index d2eff622c1b84a96bdeb4012cb0206a33012a04d..3bff11ddd6136ada45809568828cbc8f
    * - `'none'`: Disable reasoning
    */
 -  reasoningEffort: z.enum(["high", "none"]).optional()
-+  reasoningEffort: z.enum(["high", "none"]).optional(),
+  reasoningEffort: z.string().optional(),
 +  promptCacheKey: z.string().optional()
 });
 
@@ -655,7 +659,8 @@ index 54b29c08517d348995b6ca093b11160e453d5c8b..de30c3e7d924889339e38b1067cb26e9
@@ -64,6 +64,11 @@ export const mistralLanguageModelOptions = z.object({
    * - `'none'`: Disable reasoning
    */
-   reasoningEffort: z.enum(['high', 'none']).optional(),
+-  reasoningEffort: z.enum(['high', 'none']).optional(),
+  reasoningEffort: z.string().optional(),
 +
 +  /**
 +   * A stable identifier used to route requests with shared prompt prefixes.
@@ -1,8 +1,33 @@
 diff --git a/dist/index.d.mts b/dist/index.d.mts
-index 266c5ffdd9ee74ff95908ce90858ee4369d4e4ae..990ef4195bc67b6d25f249e1c81cf51710390f9a 100644
+index 266c5ffdd9ee74ff95908ce90858ee4369d4e4ae..6ac6e2873b7681ac632c903694aa38c7b09773fa 100644
 --- a/dist/index.d.mts
 +++ b/dist/index.d.mts
-@@ -78,6 +78,7 @@ declare const xaiLanguageModelResponsesOptions: z.ZodObject<{
+@@ -5,12 +5,7 @@ import { FetchFunction } from '@ai-sdk/provider-utils';
+ 
+ type XaiChatModelId = 'grok-4.3' | 'grok-4.20-0309-reasoning' | 'grok-4.20-0309-non-reasoning' | 'grok-4.20-multi-agent-0309' | 'grok-build-0.1' | (string & {});
+ declare const xaiLanguageModelChatOptions: z.ZodObject<{
+-    reasoningEffort: z.ZodOptional<z.ZodEnum<{
+-        none: "none";
+-        low: "low";
+-        medium: "medium";
+-        high: "high";
+-    }>>;
+    reasoningEffort: z.ZodOptional<z.ZodString>;
+     logprobs: z.ZodOptional<z.ZodBoolean>;
+     topLogprobs: z.ZodOptional<z.ZodNumber>;
+     parallel_function_calling: z.ZodOptional<z.ZodBoolean>;
+@@ -68,16 +63,12 @@ type XaiResponsesModelId = 'grok-4.3' | 'grok-4.20-0309-reasoning' | 'grok-4.20-
+  * @see https://docs.x.ai/docs/api-reference#create-new-response
+  */
+ declare const xaiLanguageModelResponsesOptions: z.ZodObject<{
+-    reasoningEffort: z.ZodOptional<z.ZodEnum<{
+-        none: "none";
+-        low: "low";
+-        medium: "medium";
+-        high: "high";
+-    }>>;
+    reasoningEffort: z.ZodOptional<z.ZodString>;
+     logprobs: z.ZodOptional<z.ZodBoolean>;
     topLogprobs: z.ZodOptional<z.ZodNumber>;
     store: z.ZodOptional<z.ZodBoolean>;
     previousResponseId: z.ZodOptional<z.ZodString>;
@@ -11,10 +36,35 @@ index 266c5ffdd9ee74ff95908ce90858ee4369d4e4ae..990ef4195bc67b6d25f249e1c81cf517
         "file_search_call.results": "file_search_call.results";
     }>>>>;
 diff --git a/dist/index.d.ts b/dist/index.d.ts
-index 266c5ffdd9ee74ff95908ce90858ee4369d4e4ae..990ef4195bc67b6d25f249e1c81cf51710390f9a 100644
+index 266c5ffdd9ee74ff95908ce90858ee4369d4e4ae..6ac6e2873b7681ac632c903694aa38c7b09773fa 100644
 --- a/dist/index.d.ts
 +++ b/dist/index.d.ts
-@@ -78,6 +78,7 @@ declare const xaiLanguageModelResponsesOptions: z.ZodObject<{
+@@ -5,12 +5,7 @@ import { FetchFunction } from '@ai-sdk/provider-utils';
+ 
+ type XaiChatModelId = 'grok-4.3' | 'grok-4.20-0309-reasoning' | 'grok-4.20-0309-non-reasoning' | 'grok-4.20-multi-agent-0309' | 'grok-build-0.1' | (string & {});
+ declare const xaiLanguageModelChatOptions: z.ZodObject<{
+-    reasoningEffort: z.ZodOptional<z.ZodEnum<{
+-        none: "none";
+-        low: "low";
+-        medium: "medium";
+-        high: "high";
+-    }>>;
+    reasoningEffort: z.ZodOptional<z.ZodString>;
+     logprobs: z.ZodOptional<z.ZodBoolean>;
+     topLogprobs: z.ZodOptional<z.ZodNumber>;
+     parallel_function_calling: z.ZodOptional<z.ZodBoolean>;
+@@ -68,16 +63,12 @@ type XaiResponsesModelId = 'grok-4.3' | 'grok-4.20-0309-reasoning' | 'grok-4.20-
+  * @see https://docs.x.ai/docs/api-reference#create-new-response
+  */
+ declare const xaiLanguageModelResponsesOptions: z.ZodObject<{
+-    reasoningEffort: z.ZodOptional<z.ZodEnum<{
+-        none: "none";
+-        low: "low";
+-        medium: "medium";
+-        high: "high";
+-    }>>;
+    reasoningEffort: z.ZodOptional<z.ZodString>;
+     logprobs: z.ZodOptional<z.ZodBoolean>;
     topLogprobs: z.ZodOptional<z.ZodNumber>;
     store: z.ZodOptional<z.ZodBoolean>;
     previousResponseId: z.ZodOptional<z.ZodString>;
@@ -23,9 +73,18 @@ index 266c5ffdd9ee74ff95908ce90858ee4369d4e4ae..990ef4195bc67b6d25f249e1c81cf517
         "file_search_call.results": "file_search_call.results";
     }>>>>;
 diff --git a/dist/index.js b/dist/index.js
-index 717b74538f5c8f0d6ab1475ebb2a84a47ccd3950..dd7dbeb3bc307e0d355f4bb4939d06cc8eae7528 100644
+index 717b74538f5c8f0d6ab1475ebb2a84a47ccd3950..0fd8f0d1cae951cd24401034a9c1dba762d9fd84 100644
 --- a/dist/index.js
 +++ b/dist/index.js
+@@ -246,7 +246,7 @@ var searchSourceSchema = import_v4.z.discriminatedUnion("type", [
+   rssSourceSchema
+ ]);
+ var xaiLanguageModelChatOptions = import_v4.z.object({
+-  reasoningEffort: import_v4.z.enum(["none", "low", "medium", "high"]).optional(),
+  reasoningEffort: import_v4.z.string().optional(),
+   logprobs: import_v4.z.boolean().optional(),
+   topLogprobs: import_v4.z.number().int().min(0).max(8).optional(),
+   /**
@@ -1119,6 +1119,14 @@ async function convertToXaiResponsesInput({
                   type: "input_file",
                   file_url: block.data.toString()
@@ -41,6 +100,15 @@ index 717b74538f5c8f0d6ab1475ebb2a84a47ccd3950..dd7dbeb3bc307e0d355f4bb4939d06cc
               } else {
                 throw new import_provider4.UnsupportedFunctionalityError({
                   functionality: `file part media type ${block.mediaType} as inline data (xAI Responses requires a URL or a Files API reference for non-image files)`
+@@ -1746,7 +1754,7 @@ var xaiLanguageModelResponsesOptions = import_v47.z.object({
+    * tokens), `medium` and `high` (uses more reasoning tokens). Not all models
+    * support reasoning effort; see xAI's docs for the values each model accepts.
+    */
+-  reasoningEffort: import_v47.z.enum(["none", "low", "medium", "high"]).optional(),
+  reasoningEffort: import_v47.z.string().optional(),
+   logprobs: import_v47.z.boolean().optional(),
+   topLogprobs: import_v47.z.number().int().min(0).max(8).optional(),
+   /**
@@ -1760,6 +1768,10 @@ var xaiLanguageModelResponsesOptions = import_v47.z.object({
    * The ID of the previous response from the model.
    */
@@ -63,9 +131,18 @@ index 717b74538f5c8f0d6ab1475ebb2a84a47ccd3950..dd7dbeb3bc307e0d355f4bb4939d06cc
     };
     if (xaiTools2 && xaiTools2.length > 0) {
 diff --git a/dist/index.mjs b/dist/index.mjs
-index a26af109585fc2bd3053b320142aa869c06d36f4..774adaf971b648544317a4fc65d0c56e488d4fc7 100644
+index a26af109585fc2bd3053b320142aa869c06d36f4..5faca56477b4e55a87f6f57850731c7d3e1721a5 100644
 --- a/dist/index.mjs
 +++ b/dist/index.mjs
+@@ -230,7 +230,7 @@ var searchSourceSchema = z.discriminatedUnion("type", [
+   rssSourceSchema
+ ]);
+ var xaiLanguageModelChatOptions = z.object({
+-  reasoningEffort: z.enum(["none", "low", "medium", "high"]).optional(),
+  reasoningEffort: z.string().optional(),
+   logprobs: z.boolean().optional(),
+   topLogprobs: z.number().int().min(0).max(8).optional(),
+   /**
@@ -1122,6 +1122,14 @@ async function convertToXaiResponsesInput({
                   type: "input_file",
                   file_url: block.data.toString()
@@ -81,6 +158,15 @@ index a26af109585fc2bd3053b320142aa869c06d36f4..774adaf971b648544317a4fc65d0c56e
               } else {
                 throw new UnsupportedFunctionalityError3({
                   functionality: `file part media type ${block.mediaType} as inline data (xAI Responses requires a URL or a Files API reference for non-image files)`
+@@ -1749,7 +1757,7 @@ var xaiLanguageModelResponsesOptions = z7.object({
+    * tokens), `medium` and `high` (uses more reasoning tokens). Not all models
+    * support reasoning effort; see xAI's docs for the values each model accepts.
+    */
+-  reasoningEffort: z7.enum(["none", "low", "medium", "high"]).optional(),
+  reasoningEffort: z7.string().optional(),
+   logprobs: z7.boolean().optional(),
+   topLogprobs: z7.number().int().min(0).max(8).optional(),
+   /**
@@ -1763,6 +1771,10 @@ var xaiLanguageModelResponsesOptions = z7.object({
    * The ID of the previous response from the model.
    */
@@ -158,9 +244,18 @@ index f90df62eb9a30154388b1390e9f3acc3ccc022bf..00e61cba6cf048ae0045be692f33cb7e
 
     if (xaiTools && xaiTools.length > 0) {
 diff --git a/src/responses/xai-responses-options.ts b/src/responses/xai-responses-options.ts
-index f8e96c061bf8793a402ababb8cad65bb2ad6aead..15c168892c1e8755453c61d3061e958cfd51ac71 100644
+index f8e96c061bf8793a402ababb8cad65bb2ad6aead..2a39a36221ab23ea0000bff1d7854c5bce3f9d74 100644
 --- a/src/responses/xai-responses-options.ts
 +++ b/src/responses/xai-responses-options.ts
+@@ -18,7 +18,7 @@ export const xaiLanguageModelResponsesOptions = z.object({
+    * tokens), `medium` and `high` (uses more reasoning tokens). Not all models
+    * support reasoning effort; see xAI's docs for the values each model accepts.
+    */
+-  reasoningEffort: z.enum(['none', 'low', 'medium', 'high']).optional(),
+  reasoningEffort: z.string().optional(),
+   logprobs: z.boolean().optional(),
+   topLogprobs: z.number().int().min(0).max(8).optional(),
+   /**
@@ -32,6 +32,10 @@ export const xaiLanguageModelResponsesOptions = z.object({
    * The ID of the previous response from the model.
    */
@@ -172,3 +267,16 @@ index f8e96c061bf8793a402ababb8cad65bb2ad6aead..15c168892c1e8755453c61d3061e958c
   /**
    * Specify additional output data to include in the model response.
    * Example values: 'file_search_call.results'.
+diff --git a/src/xai-chat-options.ts b/src/xai-chat-options.ts
+index d70a72a9fa01da2c711c291da5ce949efbde60b5..fd6b1ae025388b614f08b620244be553199479ca 100644
+--- a/src/xai-chat-options.ts
+++ b/src/xai-chat-options.ts
+@@ -51,7 +51,7 @@ const searchSourceSchema = z.discriminatedUnion('type', [
+ 
+ // xai-specific provider options
+ export const xaiLanguageModelChatOptions = z.object({
+-  reasoningEffort: z.enum(['none', 'low', 'medium', 'high']).optional(),
+  reasoningEffort: z.string().optional(),
+   logprobs: z.boolean().optional(),
+   topLogprobs: z.number().int().min(0).max(8).optional(),
+ 
@@ -2,7 +2,7 @@
  "name": "opencode",
  "displayName": "opencode",
  "description": "opencode for VS Code",
-  "version": "1.18.16",
+  "version": "1.18.17",
  "publisher": "sst-dev",
  "repository": {
    "type": "git",
Author	SHA1	Message	Date
opencode-agent[bot]	14b37df391	chore: update nix node_modules hashes	2026-08-12 22:36:40 +00:00
opencode-agent[bot]	91df883231	fix(opencode): select Kimi prompt by provider (#42161 ) Co-authored-by: Aiden Cline <rekram1-node@users.noreply.github.com>	2026-08-12 17:29:03 -05:00
opencode-agent[bot]	6fea419feb	fix(groq): pass through reasoning effort (#42166 ) Co-authored-by: Aiden Cline <rekram1-node@users.noreply.github.com>	2026-08-12 17:20:23 -05:00
opencode-agent[bot]	beeabe2e4b	fix(mistral): pass through reasoning effort (#42164 ) Co-authored-by: Aiden Cline <rekram1-node@users.noreply.github.com>	2026-08-12 17:20:07 -05:00
opencode-agent[bot]	502310f4df	fix(xai): pass through reasoning effort (#42160 ) Co-authored-by: Aiden Cline <rekram1-node@users.noreply.github.com>	2026-08-12 17:19:17 -05:00
opencode	37fe5c83dc	sync release versions for v1.18.17	2026-08-12 20:25:04 +00:00
Aiden Cline	dab2637217	fix(compaction): adjust instructions and structure to be more clear to smaller models like dsv4 flash (#42045 ) Co-authored-by: akenra <37288280+akenra@users.noreply.github.com>	2026-08-12 14:36:51 -05:00
opencode-agent[bot]	39fb919a05	chore: add neriousy to team members (#42107 ) Co-authored-by: Aiden Cline <rekram1-node@users.noreply.github.com>	2026-08-12 13:30:38 -05:00
opencode-agent[bot]	999be62662	chore: generate	2026-08-12 16:56:04 +00:00
opencode-agent[bot]	521906f5fa	docs(go): clarify DeepSeek ZDR coverage (#42085 ) Co-authored-by: Dax Raad <mail@thdxr.com>	2026-08-12 12:53:58 -04:00
Frank	069165eb5b	Merge branch 'dev' of github.com:anomalyco/opencode into dev	2026-08-12 12:48:10 -04:00
Frank	df09c3ec61	update ds v4 pro	2026-08-12 12:48:08 -04:00
opencode-agent[bot]	6d3ae4d63d	chore: generate	2026-08-12 16:47:08 +00:00
Adam	284187ac55	fix(ci): authenticate pulumi downloads	2026-08-12 11:45:00 -05:00
Adam	7e0353cca9	fix(stats): correct r2 daily totals	2026-08-12 11:36:23 -05:00
SKY ZHAO	959c8bd498	docs: fix provider display name and PAT typos (#42034 ) Co-authored-by: skyzhao1223 <skyzhao1223@users.noreply.github.com>	2026-08-12 10:32:23 -05:00
SKY ZHAO	ca3df21b7f	docs: fix broken DigitalOcean and Daytona links (#42048 ) Co-authored-by: skyzhao1223 <skyzhao1223@users.noreply.github.com>	2026-08-12 10:31:47 -05:00
Matthew Feroz	8571a922db	fix(provider): add Merge Gateway reasoning variants (#41867 )	2026-08-12 10:31:19 -05:00
Frank	d92d1e654b	docs(zen): add Grok 4.6	2026-08-12 10:53:38 -04:00
Adam	46a14e685a	feat(stats): query r2 data catalog	2026-08-12 09:49:10 -05:00