Claude Opus 5.5 is now live in Devin Desktop and Devin CLI.
On FrontierCode 1.1, our proprietary benchmark that grades models on real-world engineering tasks based on quality and mergeability, Claude Opus 5.5 is the new #1. It edges out Fable 5 and GPT-6 Astra while costing less than a tenth of Fable 5 and about a sixth of Astra per task.
In our evaluations, it shows particular strength on tasks where the right answer is to follow existing repo conventions. It gets to a fix with far less exploration than Fable 5, roughly 4x faster, and is disciplined about verification and keeping diffs scoped to the task.