Cognition Adds Gemini 4 Carbon to Devin, Reports Higher Coding Scores

At a glance

Cognition's AI development tool Devin now lets users run Gemini 4 Carbon in the Devin Desktop app and Devin CLI, a command-line tool. The company says the model scored higher than its predecessor on its own test of real engineering tasks.

Key points

1. 1

   Gemini 4 Carbon scored 40.0% on the FrontierCode 2.0 test.

2. 2

   That is well above the 26.3% of its predecessor, Gemini 4 Argon.

3. 3

   It also edges out GPT-6.1 Sol's 39.0% on the same test.

4. 4

   At low reasoning effort, it already beats Argon's best score.

Why it matters

It gives developers a basis for weighing both performance and cost when choosing an AI model. Lower cost per task at the same level of performance also matters.