Devin Desktop adds xAI's Grok 4.7 model

Your move: Strong on hard backend tasks, but tends to change more than asked

Cognition added xAI's newly released Grok 4.7 model to both Devin Desktop and the Devin CLI on September 21, 2026. Cognition's own testing found Grok 4.7 does best on hard, multi-module backend tasks in languages like Java, Go, and Ruby, where it digs into dependencies and checks its own work with real tests. But on Cognition's overall benchmark, it scored lower than Grok 4.6 because it tends to change more code than a task actually asks for. It costs and runs at the same speed as Grok 4.6.

Why it matters

Devin Desktop users now have another model to choose from, but Cognition's own numbers say to expect larger, less-targeted changes than you'd get from Grok 4.6. Keep an eye on the diffs it produces, especially on tasks where you want a small, contained fix.

Source: Devin blog ↗

Published by Agent Log. The summary reports the linked source; “Why it matters” and the verdict are Agent Log's interpretation.

More Devin Desktop updates →