Artificial Analysis launches FrontierCode benchmark for maintainer-merge coding tasks
Original titleAnnouncing FrontierCode on Artificial Analysis: can AI models write code maintainers would merge?
AISummary
Artificial Analysis launched FrontierCode 1.1, a coding benchmark built by Cognition with maintainers of 36 open source repositories.
Claude Opus 5.5 (max with fallback) leads with a 52% average score and a 58% pass rate, and leading models still fail more than 40% of the time.
The benchmark also penalizes excessive changes, out-of-scope file edits, and unhelpful tests, not just failing tests.
Source: Artificial Analysis Articles · artificialanalysis.aiPublished · added here