Breaking

Devin Adds Claude Sonnet 5, Topping Opus 4.8 on In-House Benchmark

June 30, 2026 at 14:23 EDT

AI Agents
Software Dev & Coding
Foundation Models

Cognition AI said on June 30, 2026, that Anthropic's Claude Sonnet 5 is now available in the Desktop and CLI versions of its autonomous coding agent Devin, with the model scoring higher than the more senior Opus 4.8 on the company's own benchmark.

June 30 · Cognition

A cheaper mid-tier model just outscored the flagship inside Devin

Claude Sonnet 5 is now the backend model in Devin Desktop and Devin CLI — and Cognition says it beat the higher-tier Opus 4.8 on its own FrontierCode benchmark, delivering frontier-level coding at a lower price.

Environments now running Sonnet 5: Devin Desktop & CLI

51.8%

Opus 4.8 score on FrontierCode Extended — the bar Sonnet 5 cleared

150

Tasks in the Extended tier measuring mergeable code quality

The Sonnet series as a "workhorse"

Gains Cognition reported when it integrated Claude Sonnet 4.5 (Sept 2025).

18%

Planning

12%

End-to-end eval

2×

Speedup

What Devin does autonomously

Plan

→

Execute & Test

→

Debug

→

Pull Request

What developers value

High token efficiency
Real-time usage display for budgeting
Strength in long-running, continuous tasks

Open concerns

Possible vendor lock-in for enterprises
Setup overhead with tools like Docker
Real-world reports still limited; verdict pending

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.