This has be mentioned before, but this problem persists for me. The model gpt-5.4-mini doesn’t work with Batch API (chat/completions). The name automatically changes from gpt-5.4-mini to gpt-5.4-mini-2026-03-17-batch. I have tried i) specifying the correct timestamp (passing gpt-5.4-mini-2026-03-17, ii) getting a new API key, iii) updating to the latest openai version==2.30.0. None of these worked.
The interesting is that gpt-5.4-mini was working fine when I first tried it 5 days ago, in the exact same setting.
1 Like
Yeah, that’s definitely confusing, especially if it was working fine just a few days ago.
The model gpt-5.4-mini-2026-03-17 is supported for the Batch API. Since you’ve already tried updating the model name, refreshing your API key, and upgrading the SDK, it sounds like you’ve covered the usual troubleshooting steps.
To help narrow this down further, could you share a few request IDs from the affected batch jobs? You can find the request ID in the response headers of your API calls.
Appreciate you sharing the detailed context.
- Sky
Sure, thanks for checking into that. Here are some request IDs: "eb35fa81-9b0e-4c21-9016-2d7b1a76a657”, “caf0e221-7d58-4479-b4f0-57e275934484”, “33400392-e51f-46ce-9f7a-59294fa39487”. The error in all these is the following: “The model `gpt-5.4-mini-2026-03-17-batch` does not exist or you do not have access to it.”