Cognition launches new SWE-2 model, Rivaling Fable 5.1 and GPT-Astra

Cognition’s SWE-2 model scores 50.0% on FrontierCode 1.1 Main1, within one point of Fable 5.1, while costing 64% less and priced at a quarter of GPT-6 Astra. It runs on a 2.8-trillion-parameter Kimi K33 base using single-run RL with linear cost penalties, triple RL environments, and quantization-aware training, cutting memory use and boosting throughput. In benchmarks, SWE-2 surpasses SWE-1.7 and Grok 4.6 on both score and cost, matches GPT-5.6 Sol and Fable 5/5.1, and lowers turn count by 58% and median edit steps from 48 to 18.