Claude Opus 4.1

Source: Anthropic Date: 2025-08-05 URL: https://www.anthropic.com/news/claude-opus-4-1

Summary

Anthropic released Claude Opus 4.1 on August 5, 2025, at Opus 4 pricing across Claude.ai, Claude Code, API, Amazon Bedrock, and Google Cloud Vertex AI. Key claims: 74.5% on SWE-bench Verified, “one standard deviation improvement over Opus 4” on junior developer benchmarks, improved research and data analysis. API identifier: claude-opus-4-1-20250805. Named integration partners: GitHub, Rakuten Group, Windsurf.

Implications

• Claude model cadence thread. Opus 4.1 is a point release — same pricing tier as Opus 4, incremental coding improvement. This is Anthropic’s pattern of iterating on the Sonnet and Opus lines between major version bumps. The cadence question is whether 4.1 is a meaningful capability jump or a “stay in the game” release while Sonnet 4.x does the volume.
• SWE-bench 74.5%. This is the benchmark that matters for coding agents and developer tools. At 74.5%, Opus 4.1 is competitive with whatever the leading coding models are in August 2025 — tracking this number against OpenAI’s equivalent is the key competitive signal.
• Junior developer benchmark. “One standard deviation improvement on junior developer benchmarks” is a specific framing — implying the model is being evaluated on the kind of tasks that could replace or augment early-career engineering work, not just expert-level coding.
• Watch: whether GitHub’s reported gains translate to Copilot integration; Windsurf and Cursor’s respective default model choices; whether Sonnet 4.5 or Opus 4.1 becomes the preferred coding model in practice.