nico1207

8 ms·

1. 1.▲
   by nico1207 2mo ago

   Did you use plan mode?
2. 2.▲
   by nico1207 2mo ago

   I just told it to count from 1 to 100, and it actually correctly counted and didn't leave out any number. I'm impressed
3. 3.▲
   by nico1207 7mo ago

   Well 5.4-pro is the more expensive and more advanced version of 5.4-thinking so why wouldn't it?
4. 4.▲
   by nico1207 7mo ago

   GPT-5.3-Codex is superior to GPT-5.4 in Terminal Bench with Codex, so not really
5. 5.▲
   by nico1207 10mo ago

   They are not even leading in Terminal-Bench... GPT 5.1-codex is better than Gemini 3 Pro
6. 6.▲
   by nico1207 2y ago

   You can just view the dataset on hugging face
7. 7.▲
   by nico1207 2y ago

   GPT-4o with image output is not yet available. So what did you even test? Dall-E 3?