Cognition releases SWE-2 near frontier coding performance at sharply lower cost
Cognition released SWE-2 for Devin, reporting a 50 percent score on FrontierCode 1.1, within one point of Anthropic's Fable 5.1, at 64 percent lower cost. The company says the model was trained with reinforcement learning at multi-trillion-parameter scale and is available in Devin Desktop and CLI.
0 comments · 0 votes
Sign in to join the discussion →
No comments yet. Start the discussion.
Why it moved the index
A lower-cost model approaching a frontier coding benchmark widens access to capable software agents and raises the pace of capability diffusion. The benchmark and cost claims are provider-reported, and the release does not establish dangerous-domain autonomy or independent real-world harm.
Assessment history
-
R1
Toward 52 · confidence 85
Adds Cognition's dated SWE-2 release, including reported frontier coding performance, lower cost, and deployment through Devin.
11 Sept 2026
Share this page
-
DoomBench assesses “Cognition releases SWE-2 near frontier coding performance at sharply lower cost” as evidence moving toward doom, with magnitude 52 and confidence 85 out of 100 in the capability gains category.
-
The DoomBench assessment of “Cognition releases SWE-2 near frontier coding performance at sharply lower cost” is based on reporting from Cognition and records the editorial rationale, source quality, attribution, and revision history.
-
DoomBench summarizes “Cognition releases SWE-2 near frontier coding performance at sharply lower cost” as follows: Cognition released SWE-2 for Devin, reporting a 50 percent score on FrontierCode 1.1, within one point of Anthropic's...
https://www.doombench.com/news/cognition-releases-swe-2-near-frontier-coding-performance-at-sharply-lower-cost-2026-09-10