It's joeover for openai and antrophic. I have been using it for 3 hours now for real work and gpt-5.1 and sonnet 4.5 (thinking) does not come close.

the token efficiency and context is also mindblowing...

it feels like I am talking to someone who can think instead of a **rider that just agrees with everything you say and then fails doing basic changes, gpt-5.1 feels particulary slow and weak in real world applications that are larger than a few dozen files.

gemini 2.5 felt really weak considering the amount of data and their proprietary TPU hardware in theory allowing them way more flexibility, but gemini 3 just works and it truly understands which is something I didn't think I'd be saying for a couple more years.