Cognition releases SWE-2, a coding model trained with cost-penalized RL
Original titleIntroducing SWE-2: Pushing the Pareto Frontier
AISummary
Cognition introduces SWE-2, a coding model post-trained from Kimi K3 that scores 50.0% on FrontierCode 1.1 Main, within one point of Fable 5.1 while costing 64% less.
The post attributes the gains to an RL algorithm that trains all reasoning-effort levels in one run, with cost penalties tuned to the base model's Pareto frontier.
SWE-2 is available starting today in Devin Desktop and CLI, with rollout to Devin Web and Fusion.
AIWhy it matters
The post explains how the cost penalty and length-weighted baseline are derived, which helps readers judge the tradeoffs in coding model post-training.
Source: Cognition Blog (Devin, Windsurf) · cognition.comPublished · added here