fix: add Anthropic models claude-opus-4-8#670
Conversation
github-actions
Bot
requested review from
Alex Z (CLowbrow),
aswink and
Caitlin Pinn (cpinn)
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
github-actions
Bot
requested review from
Erin McNulty (erin2722) and
Ken Jiang (knjiang)
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 54686e0c16
ℹ️ About Codex in GitHub
Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "Codex (@codex) review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "Codex (@codex) address that feedback".
| "grok-2-1212": ["xAI"], | ||
| "grok-vision-beta": ["xAI"], | ||
| "grok-beta": ["xAI"], | ||
| "claude-opus-4-8": ["anthropic"], |
There was a problem hiding this comment.
Exposing claude-opus-4-8 as an Anthropic model means Chat Completions requests are translated through openaiParamsToAnthropicMesssageParams, but omitUnsupportedAnthropicParams only special-cases the 4.7 model names. Anthropic documents that Opus 4.8 returns 400 for non-default temperature, top_p, or top_k, so common requests that include a temperature (for example temperature: 0.7) will be forwarded and fail unless 4.8 is added to the unsupported-params handling.
Useful? React with 👍 / 👎.
There was a problem hiding this comment.
this is referencing legacy proxy code, so not relevant
There was a problem hiding this comment.
lingua does the same thing though, so need to put up a PR to fix it there
| "input_cache_write_cost_per_mil_tokens": 6.25, | ||
| "displayName": "Claude 4.8 Opus", | ||
| "reasoning": true, | ||
| "reasoning_budget": true, |
There was a problem hiding this comment.
This marks Opus 4.8 as supporting explicit reasoning budgets, but Anthropic's Opus 4.8 docs say extended thinking budgets (thinking: {type: "enabled", budget_tokens: N}) are rejected and that adaptive thinking is the only supported thinking mode. Since the proxy translates reasoning_budget/reasoning_effort into that unsupported thinking.type: "enabled" shape, clients that rely on this metadata will send requests that Anthropic rejects.
Useful? React with 👍 / 👎.
|
testing this rn... |
2 participants
fix: add Anthropic models claude-opus-4-8
Closes #669
Source issue: #669
Summary
claude-opus-4-8claude-opus-4-82
Verified metadata
Verification notes
No LLM verification step ran — model metadata was already complete in the issue.
sync_models vs proposed update
sync_models cross-check found differences. Official provider verification was used for the applied values, and sync_models discrepancies are listed below for review.