Cognition builds the Devin autonomous software engineer and trains its own SWE family of coding models. In this corpus it first appears with SWE-1.6, which reached near-Opus 4.6 coding performance at 950 tokens per second on a hundredfold more RL compute; it later ships SWE-1.7 from an open Kimi base at 1,000 tokens per second, and Devin cracks a batch of decades-old graph conjectures in a day.
Organization
Cognition
AI lab behind the Devin coding agent and the SWE model line, whose SWE-1.6 reached near-Opus coding performance at 950 tokens per second.
Sources
Attribution and reporting linked by this record; links do not imply independent verification.
- Welcome to March 3, 2026 (theinnermostloop.substack.com)
Development timeline
Newest reporting dates first. Reporting dates may differ from event dates.
Cite or correct this record
Suggest a correction — opens a GitHub draft for you to review and submit.