AI Progress Tracker

    Kimi K3 · Kimi K3

    Moonshot · 16 Jul 2026

    Ask the Advisor about this model
    82from 5
    benchmarks

    Kimi K3 is named in reporting about agent-assisted Redis vulnerability discovery and authenticated remote-code-execution proof-of-concept development. Its release date, context window, API pricing and benchmark results are not yet confirmed by public reporting.

    context: 1Mtextimagevideo$0.24 / $12 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 questions93.1%16 Jul 2026View
    OTIS Mock AIME 2024-25Competition mathematics97.2%16 Jul 2026View
    SimpleQA VerifiedShort factual questions (hallucination resistance)50.6%27 Aug 2026View
    LMArena TextHuman preference Elo from blind chat votes (not a percentage)1475 Elo22 Oct 2026View
    LMArena WebDevHuman preference Elo for building web apps (not a percentage)1655 Elo23 Oct 2026View

    Weighted mean of 3 verified benchmark results: GPQA Diamond 93.1%, OTIS Mock AIME 2024-25 97.2%, SimpleQA Verified 50.6%.

    Capabilities

    • Reported agent-assisted discovery of Redis vulnerabilities.
    • Reported assistance constructing authenticated remote-code-execution proofs of concept.
    • Reported security research covering multiple Redis versions.
    • Reported development of exploit chains requiring specific Redis command permissions.

    What improved

    • Not yet confirmed by public reporting.

    Known flaws & limitations

    • Not yet confirmed by public reporting.

    Risks, safety & security issues

    • Reported assistance with working exploit construction illustrates the dual-use risk of security agents.
    • The reported Redis exploit chains require authentication and RESTORE permissions; they should not be characterized as unauthenticated attacks.
    • The Redis findings do not establish a vulnerability in Kimi K3 or other Kimi models.

    Good fit for

    • Evaluating agent-assisted vulnerability research in authorized test environments.
    • Human-reviewed security analysis with tightly scoped permissions.
    • Studying exploit-development assistance in isolated security laboratories.

    Poor fit for

    • Unsupervised security operations against production systems.
    • Deployments requiring established context limits, pricing or independently documented benchmark performance.
    Security & risk dashboardLatest vulnerabilities, advisories and real-world incidents for Kimi K3

    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.