Cognition 搞了个新模型 SWE-1.7，跑在 Devin 里速度超快，成本还低，值得看看。
Cognition 推出 SWE-1.7 模型，自称其训练过的最强模型。该模型在多个基准测试中得分仅比顶尖前沿模型低数个百分点，但成本大幅降低。SWE-1.7 已在 Devin 中实测并获得优异表现，推理速度可达 1000 tok/s。团队表示强化学习还有提升空间，未来将继续扩大规模。
Benchmark scores are exciting but more importantly we are seeing incredible results so far using thi...
Benchmark scores are exciting but more importantly we are seeing incredible results so far using this model in Devin! Try it out: Cognition @cognition Introducing SWE-1.7, the most capable model we’ve trained yet. It scores within a few points of the strongest frontier models at a fraction of the cost, and is now available at 1000 tok/s. RL is not hitting its limit: after refining our recipe, we keep seeing gains as we scale 🔗 View Quoted Tweet 💬 10 🔄 7 ❤️ 176 👀 11427 📊 25 ⚡