A recent evaluation of Cognition's SWE-1.6 model on 18 coding tasks revealed significant improvements over its predecessor, SWE-1.5. The new version achieved a 10-point increase in performance compared to Cognition's previous flagship model. Notably, SWE-1.6 accomplished this with fewer conversational turns and maintained the same processing speed of 950 tokens per second. AI
Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →
IMPACT Demonstrates significant performance gains in coding tasks, potentially influencing the development of future AI coding assistants.
RANK_REASON The cluster reports on an evaluation of a specific AI model's performance on coding tasks, which falls under research. [lever_c_demoted from research: ic=1 ai=1.0]
