Anthropic ships Claude Opus 4.1
Anthropic reported 74.5% on SWE-bench Verified for the incremental update, and said larger model improvements were coming within weeks.
- Models & capabilities
- Minor
Anthropic released Claude Opus 4.1, an incremental update to Opus 4 aimed at agentic tasks, real-world coding and reasoning rather than a generational jump. The company reported a score of 74.5% on SWE-bench Verified, a benchmark of real GitHub issues, and pointed to gains in research and data-analysis tasks that require tracking detail across an agentic search session.
Third parties cited in Anthropic’s announcement offered corroborating but informal evidence: GitHub reported improvements across most coding capabilities, with particular gains in multi-file refactoring; Rakuten praised the model’s precision at pinpointing exact fixes in large codebases; and Windsurf measured roughly a one-standard-deviation improvement over Opus 4 on its internal junior-developer benchmark. Opus 4.1 launched at the same price as Opus 4, across Claude’s consumer apps, the API, Claude Code, Amazon Bedrock and Google Cloud’s Vertex AI.
Anthropic framed the release as a stopgap, telling users to expect “substantially larger” improvements within weeks — a signal, confirmed when Claude Opus 4.5 and its intervening siblings shipped later that year, that Anthropic had moved to a faster release cadence of smaller, more frequent updates rather than infrequent major versions.