Cognition Unveils SWE-2 Coding Model

Cognition launched SWE-2, a coding model scoring 50.0 on FrontierCode 1.1 Main—within one point of Fable 5.1 while 64% cheaper; it was post-trained from the 2.8T-parameter Kimi K3 and uses single-run RL across reasoning levels to improve cost-performance.

• GPT-6 Astra helps Devin test software2026-09-12
• GPT-6 Astra outlasts other agents in Minecraft test2026-09-16
• Report: China narrows AI gap with U.S.2026-09-15
• Open Chinese models narrow gap with frontier models2026-09-15
• CrofAI exposed as OpenRouter wrapper with big markups2026-09-15