Devin announced that the Gemini 4 Hydrogen model is now available in its Desktop and CLI interfaces.
On the FrontierCode 2.0 benchmark, which evaluates real‑world engineering tasks, Gemini 4 Hydrogen achieved a score of 38.2%. Cognition said that places the model within a point of GPT‑6.1 Sol’s 39.0% and close to Gemini 4 Carbon’s 40.0%.
The company reported a cost of $2.40 per task, about 40% lower than Gemini 4 Carbon’s $4.04. At medium reasoning effort, the model already exceeds Gemini 4 Argon’s best result while costing less than a seventh per task.
Cognition noted that the score is a weighted aggregate of rubric items, with solutions that fail blocking criteria receiving zero. Cost figures represent the mean USD spend per rollout at list API pricing, including public discounts.
Sources
Fan coverage from Devin Central — not an official Cognition announcement. Devin is a trademark of Cognition.