Cognition Releases Fusion Harness Claiming Up To 39 Percent Cost Cut
Cognition's pitch is that harness design, not model choice alone, decides the cost of an agent run, and it is now selling that harness inside Devin Desktop and CLI.
Reporting from 1 source: GIGAZINE.
Cognition, the company behind the AI agent Devin, released Fusion, a harness that pairs a planning and review model with a cheaper execution model. Cognition advises GPT-6 Astra or Anthropic's the model as the planning model and the SWE-2 coding model as the executor, and states Fusion cuts costs by up to 39 percent versus Codex at comparable scores.
Fusion runs two agents in parallel with separate contexts and tools, exchanging only summaries, results, and feedback instead of the whole conversation. Cognition states that lets the pair make maximum use of prompt caching.
The company advises a top model for planning and review and a cheaper one for execution, naming the SWE-2 coding model, announced September 11, 2026. Fusion ships in Devin Desktop and CLI.
Synthesized by Yomimono from the 1 cited source below, including Japanese-language reporting where cited, then editorially reviewed before publishing.