Title says it all, none of the 4.1 family models are accessing the retrieval tool within assistants. Is this a planned integration, or should I remove the assistant out of the equation and just run 4.1 with the search running manually?
Thanks
Title says it all, none of the 4.1 family models are accessing the retrieval tool within assistants. Is this a planned integration, or should I remove the assistant out of the equation and just run 4.1 with the search running manually?
Thanks
I confirm that 4.1 does not use “File Search” under any settings or instructions in the “System instructions”
I’m seeing the same behavior, basically makes 4.1 not usable for me.
Same issue. Great answers, but just based on what you include right in the context. Maybe they expect to use without File Search/RAG and put your Data Source in Context. But of course, it consumes much more tokens… not a way for me.
I don’t know if this was ever fixed with the assistants API, but hopefully you’ve seen OpenAI is deprecating the Assistants API in about a year. I had this issue and have made the transition to Responses API. I can confirm 4.1 and GPT 5 both access vector search as expected in the Responses API.
Not sure, but yeah I’ve migrated to responses. It’s a little messier than assistant API but it works pretty well. GPT-5 tends to timeout if reasoning effort is set to high with multiple input types attached (images, docs, function calling) which is annoying but hey, it does the trick most of the time haha
Use background mode to avoid timeouts. It is similar to how Assistants handled polling.
This topic was automatically closed 2 days after the last reply. New replies are no longer allowed.