Cognition's FrontierCode 2.0 puts Gemini 4 Carbon at 40%, behind Claude Opus 5.5, GPT-6 Astra and Claude Sonnet 5.5

Cognition, the maker of Devin, evaluated Gemini 4 Carbon and Gemini 4 Argon on FrontierCode 2.0, its coding benchmark that measures a model's ability to go beyond the spec. These are Cognition's own measurements.

• Gemini 4 Carbon: 40% at $4.04 per task, which Cognition calls near state of the art, behind Claude Opus 5.5, GPT-6 Astra and Claude Sonnet 5.5.
• Gemini 4 Argon: 26.3% at $6.41 per task, ahead of Muse Wonder 1.0 but behind Claude Haiku 5.5 and Grok 4.7.

Full scores are on Cognition's FrontierCode leaderboard. One reply on X asks why Devin CLI was used as the harness for the Gemini models rather than Antigravity.