Does anyone know if the new model o1-preview is compatible with OPENAI’s assistants. I want to try it out but I dont see it as available in the playground for Assistants.
If it is, is it preforming well for you?
I don’t see it in Assistants / Playground either. I can use it in the chat though.
Tested it with ~30 prompt in different projects on Knit, it doesn’t outperform much compared with the current 4-o or other models in these tasks. But that doesn’t mean it’s not good, I think we need to explore new scenarios with the capability of O1 models.
The technical writers at OpenAI know:
Both
o1-previewando1-minican be accessed using the chat completions endpoint.
Assistants makes heavy use of an instruction (system message), tool calls, and developer functions to deliver its features, not currently supported.
2 Likes
I found it outperforms if you have o1 explain how to solve a complex problem and then have 4o or even 4o mini complete the steps.
Seems that the reasoning helps it assist other models and get noticeably better accuracy.
1 Like
Similar. o1 seem to be building my python parsers more accurately than GPT-4o. I’m finding I have to intervene less frequently for sure.
When will it be available for Building Assistants, any updates?
1 Like