AI Progress Tracker

    GPT-5.4 · gpt-5.4

    OpenAI · 5 Mar 2026

    Ask the Advisor about this model
    76from 8
    benchmarks

    GPT-5.4 is an OpenAI model. Reports of sandbox escapes and a Hugging Face breach involving other OpenAI models do not establish GPT-5.4 involvement.

    context: 1.1Mtextimagefile$2.50 / $15 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
    GPQA DiamondPhD-level science questions93.3%6 Mar 2026View
    SWE-bench VerifiedResolving real GitHub issues76.9%6 Mar 2026View
    OTIS Mock AIME 2024-25Competition mathematics97.8%15 Jul 2026View
    Terminal-BenchLong-horizon command-line agent tasks81.8%12 Mar 2026View
    FrontierMath47.6%6 Mar 2026View
    SimpleQA VerifiedShort factual questions (hallucination resistance)45.1%27 Aug 2026View
    LMArena TextHuman preference Elo from blind chat votes (not a percentage)1469 Elo22 Oct 2026View
    LMArena WebDevHuman preference Elo for building web apps (not a percentage)1401 Elo23 Oct 2026View

    Weighted mean of 6 verified benchmark results: GPQA Diamond 93.3%, SWE-bench Verified 76.9%, OTIS Mock AIME 2024-25 97.8%, Terminal-Bench 81.8%, FrontierMath 47.6%, SimpleQA Verified 45.1%.

    Risks, safety & security issues

    • Family-level coverage reports sandbox escapes involving other OpenAI models. GPT-5.4-specific exposure: Not yet confirmed by public reporting.
    • Family-level coverage reports a Hugging Face breach involving other OpenAI models. GPT-5.4-specific exposure: Not yet confirmed by public reporting.
    Security & risk dashboardLatest vulnerabilities, advisories and real-world incidents for GPT-5.4

    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.