AI Progress Tracker

    Claude Opus 4.1 · claude-opus-4-1-20250805

    Anthropic · 5 Aug 2025

    Ask the Advisor about this model
    57from 7
    benchmarks

    Claude Opus 4.1 is Anthropic’s August 2025 model for coding, agentic workflows, research and analysis, with a documented 200K-token context window. Its published SWE-bench Verified result is 74.5%, and newer Opus generations, including Opus 4.5, have subsequently shipped.

    context: 200Kimagetextfile$15 / $75 per 1M tokens (in / out)

    Benchmark results

    Every figure links to where it was published. Compare like with like — harnesses and prompt setups differ between vendors.

    BenchmarkScoreSource
    SWE-bench VerifiedResolving real GitHub issues74.5%5 Aug 2025View
    GPQA DiamondPhD-level science questions77.3%5 Aug 2025View
    OTIS Mock AIME 2024-25Competition mathematics68.9%5 Aug 2025View
    Terminal-BenchLong-horizon command-line agent tasks38%31 Oct 2025View
    FrontierMath7.2%5 Aug 2025View
    LMArena TextHuman preference Elo from blind chat votes (not a percentage)1419 Elo22 Oct 2026View
    LMArena WebDevHuman preference Elo for building web apps (not a percentage)1391 Elo23 Oct 2026View

    Weighted mean of 5 verified benchmark results: SWE-bench Verified 74.5%, GPQA Diamond 77.3%, OTIS Mock AIME 2024-25 68.9%, Terminal-Bench 38%, FrontierMath 7.2%.

    Capabilities

    • Software engineering, including repository-level debugging and multi-file code changes.
    • Agentic workflows involving tools and multiple execution steps.
    • Research and data analysis with improved detail tracking.
    • Extended thinking for complex reasoning tasks.
    • Text and image input with text output.

    What improved

    • Anthropic’s SWE-bench Verified result increased from 72.5% for Opus 4 to 74.5% for Opus 4.1.
    • GitHub reported stronger multi-file code refactoring than Opus 4.
    • Rakuten reported more precise debugging corrections with fewer unnecessary changes.
    • Anthropic reported improvements in research, data analysis, detail tracking and agentic search.

    Known flaws & limitations

    • The documented context window is 200K tokens.
    • The documented maximum output is 32K tokens.
    • Native audio and video input are not supported.
    • The published SWE-bench Verified result does not establish reliable completion of every software-engineering task.

    Risks, safety & security issues

    • Coding and tool-use capabilities can be abused to support data theft and extortion, as documented in Anthropic’s reporting on Claude Code misuse.
    • Broad access to sensitive files and external tools can increase the impact of malicious or unauthorized workflows.
    • Automated analysis of stolen information and generation of personalized extortion demands can increase the scale of criminal operations.

    Good fit for

    • Complex codebase debugging and multi-file refactoring with human review.
    • Research and document analysis within the documented context limit.
    • Tool-assisted workflows with narrowly scoped permissions and approval gates.
    • Detailed analytical tasks that benefit from extended thinking.

    Poor fit for

    • Applications requiring native audio or video understanding.
    • Tasks requiring more than 200K tokens in one context or more than 32K output tokens.
    • Unsupervised operations involving sensitive data, production systems or irreversible actions.
    Security & risk dashboardLatest vulnerabilities, advisories and real-world incidents for Claude Opus 4.1

    References

    Profile last reviewed 10/27/2026 and re-checked automatically every few hours against the vendor's published documentation, papers and public leaderboards. Scores come from the cited benchmark results above — confirm limits and pricing against the provider's own documentation before building on them.