Windsurf: Claude Sonnet 5.5 Now Available in Devin
WindsurfWindsurf's Devin added Claude Sonnet 5.5, which scores 64.4% on FrontierCode 1.1 and edges past Fable 5.1 at maximum reasoning effort. Anthropic reports over 30% faster output generation and up to 30% lower cost per task than Sonnet 5, with token pricing unchanged. The model can be tried through Devin Desktop or the Devin CLI.
Key Takeaways
- Claude Sonnet 5.5 scores 64.4% on FrontierCode 1.1, beating Fable 5.1 at maximum reasoning effort.
- Anthropic reports more than 30% faster output generation compared with Sonnet 5.
- Cost per task is up to 30% lower than Sonnet 5, even though token pricing is unchanged.
- The gain is a drop-in upgrade: no price change means existing Sonnet users benefit automatically after switching models.
- A Sonnet-tier model topping a higher-tier model on a real-world coding benchmark shifts the cost versus quality tradeoff for everyday tasks.
- The model can be tried via Devin Desktop or the Devin CLI.
Claude Sonnet 5.5 Joins Devin
Windsurf's Devin announced on September 28, 2026 that Claude Sonnet 5.5 is now available. The model is an incremental upgrade to Sonnet 5 and lands as one of the strongest performers on Cognition's FrontierCode 1.1 benchmark, which grades models on real-world coding tasks. Developers can try it by downloading Devin Desktop or installing the Devin CLI.
Benchmark Results
On FrontierCode 1.1, Claude Sonnet 5.5 scores 64.4%. Cognition notes that this exceeds even Fable 5.1 at its maximum reasoning effort, which is notable for a model in the Sonnet tier that is normally positioned as the more affordable, everyday option. The result puts Sonnet 5.5 near the top of the leaderboard for code quality.
Faster and Cheaper Than Sonnet 5
The post highlights Anthropic's own figures for the generational change. Sonnet 5.5 delivers more than 30% faster output generation and up to 30% lower cost per task compared with Sonnet 5, and token pricing is unchanged. In practice, the savings come from the model needing less work to finish a task, so existing users see lower spend and quicker responses without any change to the per-token rate card.
Who Benefits
Anyone who already uses Sonnet-class models in Devin for daily coding gets a drop-in improvement in speed and efficiency. Teams that reserved heavier, pricier models for hard problems may find Sonnet 5.5 strong enough to handle more of their workload, since it now rivals those models on FrontierCode.