Trouble with 4.1 and RAG. other models are fine

[holgerweberschwerte]

i created an assistant with the dashboard. in playground and in my own gui 4.1 doesnt seem to acknowledge my RAG document. i get good answers from 4.1mini, 4.1nano and older models. 4.1 quotes from an imaginary document. is it a known issue? does 4.1 not work properly for others? thanks in advance

[vb]

Hi and welcome to the community!

I have started to collect these reports in another topic and will move your post there for additional exposure.

Assistant API file search error after GPT-4.1 release

After GPT-4.1 was released, we transitioned 70k of our Assistants from GPT-4o to GPT-4.1. Our use case heavily relies on the Assistant API combined with RAG. Since making this update, we’ve encountered significant issues: Assistants are responding with irrelevant content, and the original source citations are no longer being provided. Even after reverting from GPT-4.1 back to GPT-4o, the problem persists—our Assistants still fail to correctly reference and respond based on RAG data. Is there …