3 ms·
GPT 5.1 / Codex already beats Gemini 3 on SWE Bench Verified and Terminal Bench and this pushes the gap further. Seems like a decent improvement.
by Palmik 11mo ago
GPT 5.1 / Codex already beats Gemini 3 on SWE Bench Verified and Terminal Bench and this pushes the gap further. Seems like a decent improvement.
- knowriju 11mo agoWould it be fair to compare a generic model with a model finetuned for coding?
- skhameneh 11mo agoThere’s been community commentary that many of the GPT models are a tad overfitted WRT benchmarks. Benchmarks are not representative of end user experiences. That’s not to say the benchmarks aren’t useful at all, but are only useful as a subjective indicator.