Gemini 4 Carbon is now available in Devin

Devin added Gemini 4 Carbon to Devin Desktop and Devin CLI.

On Devin's own FrontierCode 2.0 benchmark, which scores models on real-world engineering tasks for quality and mergeability, Gemini 4 Carbon gets 40.0%. Gemini 4 Argon scored 26.3%, and GPT-6.1 Sol scored 39.0%. Solutions that don't pass blocking criteria get a 0.

Devin also says it costs less. At low reasoning effort, Carbon already beats Argon's best result, and its top score costs less than two-thirds as much per task as Argon's. Cost is the mean USD spend per rollout at list API pricing, including public discounts. These numbers come from Devin's benchmark, not an independent test.