Cognition announced Fusion for Devin CLI and Devin Desktop on September 11, saying the dual-model setup is “up to 39% more efficient compared to other model harnesses across major coding benchmarks.” The company announced the release in a post on X.
Fusion lets users choose a frontier lead model for planning and review, then assign a lower-cost sidekick model to execution. Cognition recommends Fable 5.1 or GPT-6 Astra as the lead model, with SWE-2 as the sidekick.
Cognition’s Fusion blog post lists Codex at 100 ($7.47) and Devin Fusion at 61 ($4.54), which it describes as a 39% lower cost. Cognition published the benchmark results in its Fusion announcement.
The release follows Cognition’s SWE-2 model announcement by one day. Cognition’s blog lists SWE-2 on September 10 and Fusion in Devin Desktop and CLI on September 11. Go Big News reported on September 11 that Cognition had released SWE-2.
Cognition describes Devin as an autonomous software engineer that can plan, write, test and ship production code inside a team’s existing codebase and tools. The company was founded in August 2023 by Scott Wu, Steven Hao and Walden Yan, former competitive programmers who earned gold medals at the International Olympiad in Informatics.
