anomalyco/opencode
1
0
Fork 0
mirror of https://github.com/anomalyco/opencode.git synced 2026-07-22 02:05:26 -04:00
Code Issues 5.9k Packages Projects Releases 19 Wiki Activity

OUTPUT_TOKEN_MAX hardcoded to 32k prevents using full 64k output limit for Claude Sonnet 4.5 with extended thinking #1937

New Issue
Closed
opened 2026-02-16 17:33:19 -05:00 by yindo · 2 comments
yindo commented 2026-02-16 17:33:19 -05:00
Owner
Copy Link
Copy Source

Originally created by @calvinalkan on GitHub (Oct 3, 2025).

Originally assigned to: @rekram1-node on GitHub.

Problem

Claude Sonnet 4.5 supports 64k output tokens, but OpenCode has a hardcoded limit
that prevents using the full capacity when extended thinking is enabled.

Current Behavior

In src/session/prompt.ts:

export const OUTPUT_TOKEN_MAX = 32_000

const outputLimit = Math.min(model.info.limit.output, OUTPUT_TOKEN_MAX) ||
OUTPUT_TOKEN_MAX

This caps outputLimit at 32k even when the model config specifies 64k:

"claude-sonnet-4-20250514": {
  "limit": {
    "output": 64000,
    "context": 200000
  }
}

When using extended thinking with a 32k budget:

thinking:
  type: enabled
  budgetTokens: 32000

The calculation becomes: 32000 - 32000 = 0, causing:

AI_InvalidArgumentError: maxOutputTokens must be >= 1

Expected Behavior

OUTPUT_TOKEN_MAX should be at least 64k to support Claude Sonnet 4.5's full output
capacity with extended thinking enabled.

Workaround

Reduce thinking budgets to fit within 32k total (e.g., budgetTokens: 20000 leaves
12k for output).

Originally created by @calvinalkan on GitHub (Oct 3, 2025). Originally assigned to: @rekram1-node on GitHub. ## Problem Claude Sonnet 4.5 supports 64k output tokens, but OpenCode has a hardcoded limit that prevents using the full capacity when extended thinking is enabled. ## Current Behavior In `src/session/prompt.ts`: ```typescript export const OUTPUT_TOKEN_MAX = 32_000 const outputLimit = Math.min(model.info.limit.output, OUTPUT_TOKEN_MAX) || OUTPUT_TOKEN_MAX ``` This caps outputLimit at 32k even when the model config specifies 64k: ```json "claude-sonnet-4-20250514": { "limit": { "output": 64000, "context": 200000 } } ```` When using extended thinking with a 32k budget: ```yaml thinking: type: enabled budgetTokens: 32000 ``` The calculation becomes: 32000 - 32000 = 0, causing: AI_InvalidArgumentError: maxOutputTokens must be >= 1 ## Expected Behavior OUTPUT_TOKEN_MAX should be at least 64k to support [Claude Sonnet 4.5's full output](https://docs.claude.com/en/docs/about-claude/models/overview) capacity with extended thinking enabled. ## Workaround Reduce thinking budgets to fit within 32k total (e.g., budgetTokens: 20000 leaves 12k for output).
yindo closed this issue 2026-02-16 17:33:19 -05:00
yindo commented 2026-02-16 17:33:20 -05:00
Author
Owner
Copy Link
Copy Source

@rekram1-node commented on GitHub (Oct 3, 2025):

hm we will need to adjust out logic for the budget tokens calculation...

@rekram1-node commented on GitHub (Oct 3, 2025): hm we will need to adjust out logic for the budget tokens calculation...
yindo commented 2026-02-16 17:33:20 -05:00
Author
Owner
Copy Link
Copy Source

@rekram1-node commented on GitHub (Oct 5, 2025):

should be fixed in next release

@rekram1-node commented on GitHub (Oct 5, 2025): should be fixed in next release
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
acp
bad
bug
discussion
docs
documentation
enhancement
extension
good first issue
growth
help-wanted
mcp
model-problem
needs:compliance
nix
opentui
perf
pull-request
Mirrored from GitHub Pull Request
 tui
web
windows
zen
No Label
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
yindo
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: anomalyco/opencode#1937
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?