Chris Messina

GPT-5.1 Instant and Thinking - Smarter, more conversational ChatGPT with new personalities

GPT-5.1 Instant and Thinking are trained using the same stack as OpenAI's reasoning models, so they score higher on factuality and complex problem-solving than GPT-5, while also introducing a more natural, conversational tone. We think many people will find that GPT-5.1 does a better job of bringing IQ and EQ together, but one default clearly can’t meet everyone’s needs.

Add a comment

Replies

Best
Paul Tseluyko

Who else noticed how GPT 5 API became slow (like 2min per response)? 😬

Iren Loseva
Работаю с нейросетями каждый день. Новый тон и логика 5.1 - заметно ближе к реальной коммуникации ,буду тестировать в обучении и в контенте 🚀
Sudhir Balaji

exciting! do we have any rough timeline for API access (if at all)?

Tristan Pollock

love

Joanna Sherlin
ahhh so I was right to think it's 5.1!!!
jack cheng

If one day it creates fundamental knowledge and can prove itself to be correct, how can such issues be regulated?

Aarush Mishra
There is no assistant anymore, apis have no control model can randomly go in thinking mode that can minimum take upto 14 seconds so for getting any llm response should we keep the thread open but that would expose our apis, so what’s the solution?

Impressive update! GPT-5.1 feels more consistent in complex instructions. Any plans for exposing more granular control over the Thinking mode in the API?

Khurram Hanif

Slowdowns are noticeable. Hopefully, they address them soon.

Altaf Shaikh

Has anyone else noticed how the GPT-5 API has become slow, taking around two minutes to generate a response?