Cognition launches SWE-2 Flash, beating US open-weight models 3x its size after 6 weeks
simonguozirui · x · 2026-11-05
Cognition introduced SWE-2 Flash, a fast and efficient coding model trained in just 6 weeks — its first model with a midtraining stage combined with multiple RL improvements, delivering the largest relative gains over a base model the team has achieved.
- On leading evals it outperforms all American open-weight models, even ones 3x larger (chart: 20.0 on Terminal-Bench 4.0, 60.1 on DeepSWE 1.1, 24.0 on FrontierCode 2.0).
- The team wants to offer open-weight benefits without cyber risk (citing North Korean hackers' Binance heist as the largest theft in history), and is building a secure weight-sharing program for enterprises, partners and research groups — a compromise between transparency and safety.
Related event: Cognition releases SWE-2 Flash coding model(4 posts)→
More from coding & agent
- POKI open-sourced: a no-jailbreak iOS agent that can fully control your iPhone — alexcovo_eth · 2026-11-05
- Foresight-SWE benchmark: coding agents solve 60.7% of issues but only 14.8% without introducing bugs — OfirPress · 2026-11-05
- User: Every Client Deck Automatically Improves HQForWork's Skill File — jacob_posel · 2026-11-05
- 10 Grok Bots to Build Your First $1K: Idea Stress Tests, Competitor Stalking & More — ChrisUniverse · 2026-11-05
- Developer says Claude and Codex nearly bankrupted him by making procrastination comfortable — oran_ge · 2026-11-05
- "We're gardening now": devs rebrand their role as agent babysitters — heyneighbor · 2026-11-05