chore: sync new models#979
Conversation
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
![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: 8684fa19c0
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".
| "max_input_tokens": 128000, | ||
| "max_output_tokens": 272000, |
There was a problem hiding this comment.
For gpt-5-pro requests, this swaps the documented limits: Microsoft Foundry lists gpt-5-pro (2025-10-06) as 272,000 input / 128,000 output (https://learn.microsoft.com/en-us/azure/foundry/foundry-models/concepts/models-sold-directly-by-azure#gpt-5). With 128,000 input and 272,000 output in the catalog, clients will unnecessarily avoid valid >128k prompts and may advertise or try 272k completions that OpenAI/Azure reject; the dated alias below should be fixed with the same values.
Useful? React with 👍 / 👎.
This "sync new models" run reverted every recent manual correction because the SYNC_PRESERVED_FIELDS pins are not merged yet: - gpt-5 / gpt-5-mini / gpt-5-pro (+ snapshots) + gpt-5.3-codex max_input 400000 -> 272000/128000 (OpenAI's 400k window) - gpt-5.4-pro / gpt-5.5-pro (+ snapshots) re-added a fake $3/M cache-read (Pro tier has no cached pricing) - grok-code-fast -> old $0.20/$0.02/$1.50 (xAI prices it as grok-build-0.1: $1/$0.20/$2) - fireworks glm-5p2 cached input 0.14 -> 0.26 (0.26 is the GLM 5.1 rate) Revert all to main's verified values. Keep the one genuinely new model, thinkingmachines/inkling (baseten), validated 200 both modes via the gateway. Add SYNC_PRESERVED_FIELDS pins for all of the above so the sync stops reverting them (+ isFieldManuallyPreserved tests). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
1 participant
Automated sync of model metadata and newly available models from the approved provider allowlist.