Cognition Releases SWE-2: A Kimi K3 Post-Trained Coding Model That Matches Fable 5.1 on FrontierCode at 64% Lower Cost

Cognition has released SWE-2, a new coding model post-trained using reinforcement learning from the Kimi K3 model. The company reports that SWE-2 achieves a 50.0% score on the FrontierCode 1.1 Main benchmark, performing near parity with Fable 5.1 while requiring 64% less cost to operate.