anomalyco/opencode
1
0
Fork 0
mirror of https://github.com/anomalyco/opencode.git synced 2026-07-22 10:15:31 -04:00
Code Issues 5.9k Packages Projects Releases 19 Wiki Activity

[PR #4256] fix: correct cache cost for OpenRouter and other OpenAI-compatible providers #10918

New Issue
Closed
opened 2026-02-16 18:15:40 -05:00 by yindo · 0 comments
yindo commented 2026-02-16 18:15:40 -05:00
Owner
Copy Link
Copy Source

Original Pull Request: https://github.com/anomalyco/opencode/pull/4256

State: closed
Merged: Yes

fixes #534, #454, #2020

Problem: There's a significant mismatch in costs displayed in OpenCode vs OpenRouter.

OpenRouter provides inputTokens as the total count of input tokens (including cached).
AFAIK other providers (OpenRouter/OpenAI/Gemini etc.) do it the same way e.g. https://github.com/vercel/ai/issues/8794#issuecomment-3320975344

Anthropic does it differently though - inputTokens doesn't include cached tokens.
It looks like OpenCode's cost calculation assumes all providers return inputTokens the same way Anthropic does (I'm guessing getUsage logic was originally implemented with anthropic), so it's causing incorrect cost calculation for OpenRouter and others.

Testing:
Verified against OpenRouter data - costs are now matching almost perfectly.
Why almost - cache.write costs are not included in OpenRouter's metadata, so there may be minor discrepancies for the messages with cache writes. This seems to be an upstream limitation that needs to be addressed by ai-sdk/openrouter.

**Original Pull Request:** https://github.com/anomalyco/opencode/pull/4256 **State:** closed **Merged:** Yes --- fixes #534, #454, #2020 Problem: There's a significant mismatch in costs displayed in OpenCode vs OpenRouter. OpenRouter provides `inputTokens` as the total count of input tokens (including cached). AFAIK other providers (OpenRouter/OpenAI/Gemini etc.) do it the same way e.g. https://github.com/vercel/ai/issues/8794#issuecomment-3320975344 Anthropic does it differently though - `inputTokens` doesn't include cached tokens. It looks like OpenCode's cost calculation assumes all providers return `inputTokens` the same way Anthropic does (I'm guessing `getUsage` logic was originally implemented with anthropic), so it's causing incorrect cost calculation for OpenRouter and others. Testing: Verified against OpenRouter data - costs are now matching _almost_ perfectly. Why almost - `cache.write` costs are not included in OpenRouter's metadata, so there may be minor discrepancies for the messages with cache writes. This seems to be an upstream limitation that needs to be addressed by ai-sdk/openrouter.
yindo added the pull-request label 2026-02-16 18:15:40 -05:00
yindo closed this issue 2026-02-16 18:15:40 -05:00
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 pull-request
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#10918
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?