I’ve been following the cookbook to set up my batching API task. Using the same dataset and code, but switching between different GPT models (e.g., GPT-4o and GPT-4.1-mini), I’ve noticed a significant difference in response times. In particular, when using GPT-4.1-mini, the program continues running but doesn’t return any responses. Additionally, when I check my usage on the account page, I don’t see any token consumption recorded for GPT-4.1-mini. any one can help or any suggestion?
This is normal. Batches are returned within 24 hours, and this ability for OpenAI to flex compute is what allows them to offer that sweet discount for batched processing. You’ll likely start seeing progress being made at night when nobody is using AI. But as long as it hasn’t been 24 hours yet, there isn’t anything to be concerned about.
1 Like
Hi, OnceAndTwice, thank you for your replies. So… the only thing I can do is to wait for 24 hours and see whether I can get the results?
Yes, the results of your batch are only guaranteed after 24 hours, although most of the time it’ll finish before then.
The OpenAI docs has more info on batches if you’re interested in reading: https://platform.openai.com/docs/guides/batch
Learn how to use OpenAI’s Batch API to send asynchronous groups of requests with 50% lower costs, a separate pool of significantly higher rate limits, and a clear 24-hour turnaround time. The service is ideal for processing jobs that don’t require immediate responses.
Is there any place I can see the status of every GPT model’ status? I have tried several models, except GPT-4.1-mini, all other models work…
You should be able to see all your batches and their progress at https://platform.openai.com/batches
1 Like
This is not you. This is about the service level that is fulfilled by OpenAI. GPT-4.1 mini batches seem to be stalling or timing out.
Remember, OpenAI has 24 hours to service the job – or to cancel it if they don’t want to run it.
1 Like
This link helps me a lot! thank you!
1 Like