AI Progress Tracker

    GLM 4.7 · glm-4.7

    Z.ai · 22 Dec 2025

    Ask the Advisor about this model
    60from 11
    benchmarks

    GLM-4.7 is Z.ai’s text-focused model for coding, reasoning and tool-assisted workflows, with a documented 200K-token context window. Its published evaluations show strengths in software engineering and competitive mathematics, while difficult expert-level questions and autonomous code changes remain important limitations.

    context: 200Ktext$0.60 / $2.20 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
    AIMECompetition mathematics95.7%Dec 2025View
    GPQA DiamondPhD-level science questions85.7%Dec 2025View
    Humanity's Last ExamFrontier expert questions across fields24.8%Dec 2025View
    LiveCodeBenchFresh competitive-programming problems84.9%Dec 2025View
    SWE-bench VerifiedResolving real GitHub issues73.8%Dec 2025View
    OTIS Mock AIME 2024-25Competition mathematics83.3%29 Jan 2026View
    Terminal-BenchLong-horizon command-line agent tasks33.4%22 Dec 2025View
    FrontierMath2.4%30 Jan 2026View
    SimpleQA VerifiedShort factual questions (hallucination resistance)32.2%27 Aug 2026View
    LMArena TextHuman preference Elo from blind chat votes (not a percentage)1435 Elo22 Oct 2026View
    LMArena WebDevHuman preference Elo for building web apps (not a percentage)1435 Elo23 Oct 2026View

    Weighted mean of 9 verified benchmark results: AIME 95.7%, GPQA Diamond 85.7%, Humanity's Last Exam 24.8%, LiveCodeBench 84.9%, SWE-bench Verified 73.8%, OTIS Mock AIME 2024-25 83.3%, Terminal-Bench 33.4%, FrontierMath 2.4%, SimpleQA Verified 32.2%.

    Capabilities

    • Code generation, debugging and repository-level software-engineering assistance.
    • Competitive mathematics and scientific question answering.
    • Frontend implementation and interface generation.
    • Tool-assisted workflows with reasoning across multiple steps.
    • Long-context text processing within a documented 200K-token context window.

    What improved

    • Higher vendor-reported software-engineering benchmark performance than GLM-4.6.
    • Improved multilingual coding performance.
    • Improved frontend design and generated interface quality.
    • Stronger reported performance on difficult reasoning evaluations.
    • Expanded thinking controls, including preserved reasoning across tool calls and turn-level thinking configuration.

    Known flaws & limitations

    • The base model is text-focused rather than a native image, audio or video understanding model.
    • Published software-engineering results leave unresolved tasks; generated patches still require testing and review.
    • Published Humanity's Last Exam results show substantial limitations on exceptionally difficult expert-level questions.

    Risks, safety & security issues

    • Not yet confirmed by public reporting.

    Good fit for

    • Developer-supervised coding, debugging and repository maintenance.
    • Frontend prototyping with human review of usability and correctness.
    • Mathematical and scientific problem-solving assistance with independent verification.
    • Long-context, tool-assisted text workflows.

    Poor fit for

    • Native image, audio or video understanding without a separate multimodal model.
    • Unreviewed production code changes or autonomous deployment.
    • High-stakes expert decisions without qualified human verification.
    Security & risk dashboardLatest vulnerabilities, advisories and real-world incidents for GLM 4.7

    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.