{
  "api_version": "1",
  "updated": "2026-10-26",
  "latest_event": "2026-10-24",
  "rights": "© Frontier Log, all rights reserved.",
  "model": {
    "slug": "kimi-k2-thinking",
    "name": "Kimi K2 Thinking",
    "lab": "moonshot",
    "date": "2025-11-06",
    "flagship": false,
    "event": "https://frontierlog.ai/event/2025-11-06-kimi-k2-thinking",
    "releases": [],
    "variant_of": null,
    "made_by": "Moonshot AI",
    "url": "https://frontierlog.ai/model/kimi-k2-thinking",
    "results": [
      {
        "id": "kimi-k2-thinking:hle:no-tools:1",
        "model": "kimi-k2-thinking",
        "model_name": "Kimi K2 Thinking",
        "model_date": "2025-11-06",
        "event": "https://frontierlog.ai/event/2025-11-06-kimi-k2-thinking",
        "benchmark": "hle",
        "benchmark_name": "Humanity's Last Exam",
        "setting": "no-tools",
        "setting_name": "No tools",
        "score": 23.9,
        "unit": "%",
        "printed": "23.9",
        "reported_by": "moonshot",
        "run_by": null,
        "label": "K2 Thinking",
        "source": "https://www.kimi.ai/blog/kimi-k2-thinking",
        "source_title": "Introducing Kimi K2 Thinking",
        "source_date": "2025-11-06",
        "where": "'Full Evaluations' table, row 'Humanity's Last Exam (Text-only)', sub-row 'no tools', column 'K2 Thinking'",
        "own": true,
        "primary": true,
        "parallel": false,
        "notes": "text-only subset; 96K thinking-token budget; INT4 precision",
        "caveat": "",
        "human_verified": true,
        "human_verified_on": "2026-10-17",
        "check": "human",
        "corroborated": true,
        "corroborated_on": "2026-10-19",
        "corroborated_by": [
          {
            "url": "https://api-docs.deepseek.com/news/news251201/",
            "name": "DeepSeek-V3.2",
            "title": "DeepSeek-V3.2 Release",
            "lab": "deepseek",
            "date": "2025-12-01",
            "printed": "23.9",
            "where": "Reasoning table image (token counts in parentheses omitted); row 'HLE'; column 'Kimi-K2 Thinking'"
          },
          {
            "url": "https://z.ai/blog/glm-4.7",
            "name": "GLM-4.7",
            "title": "GLM-4.7: Advancing the Coding Capability",
            "lab": "zai",
            "date": "2025-12-22",
            "printed": "23.9",
            "where": "Benchmark table, row 'HLE (separate from 'HLE w/ Tools' row)', column 'Kimi K2 Thinking'"
          }
        ],
        "source_capture": "https://frontierlog.ai/captures/4e2626572b0cef3e.png"
      },
      {
        "id": "kimi-k2-thinking:browsecomp:default:1",
        "model": "kimi-k2-thinking",
        "model_name": "Kimi K2 Thinking",
        "model_date": "2025-11-06",
        "event": "https://frontierlog.ai/event/2025-11-06-kimi-k2-thinking",
        "benchmark": "browsecomp",
        "benchmark_name": "BrowseComp",
        "setting": "default",
        "setting_name": "",
        "score": 60.2,
        "unit": "%",
        "printed": "60.2",
        "reported_by": "moonshot",
        "run_by": null,
        "label": "K2 Thinking",
        "source": "https://www.kimi.ai/blog/kimi-k2-thinking",
        "source_title": "Introducing Kimi K2 Thinking",
        "source_date": "2025-11-06",
        "where": "'Full Evaluations' table, row 'BrowseComp' (Agentic Search Tasks), 'w/ tools', column 'K2 Thinking'",
        "own": true,
        "primary": true,
        "parallel": false,
        "notes": "search, code and browsing tools; up to 300 steps; older tool outputs hidden past 256K tokens",
        "caveat": "",
        "human_verified": true,
        "human_verified_on": "2026-10-17",
        "check": "human",
        "corroborated": true,
        "corroborated_on": "2026-10-19",
        "corroborated_by": [
          {
            "url": "https://www.minimax.io/news/minimax-m21",
            "name": "MiniMax M2.1",
            "title": "MiniMax M2.1: Significantly Enhanced Multi-Language Programming, Built for Real-World Complex Tasks",
            "lab": "minimax",
            "date": "2025-12-23",
            "printed": "60.2",
            "where": "Tool-use and intelligence table image; row 'BrowseComp (context management)'; column 'Kimi K2 Thinking'; with context management"
          },
          {
            "url": "https://z.ai/blog/glm-4.7",
            "name": "GLM-4.7",
            "title": "GLM-4.7: Advancing the Coding Capability",
            "lab": "zai",
            "date": "2025-12-22",
            "printed": "60.2",
            "where": "Benchmark table, row 'BrowseComp w/ Context Manage', column 'Kimi K2 Thinking'"
          }
        ],
        "source_capture": "https://frontierlog.ai/captures/4e2626572b0cef3e.png"
      },
      {
        "id": "kimi-k2-thinking:hle:with-tools:1",
        "model": "kimi-k2-thinking",
        "model_name": "Kimi K2 Thinking",
        "model_date": "2025-11-06",
        "event": "https://frontierlog.ai/event/2025-11-06-kimi-k2-thinking",
        "benchmark": "hle",
        "benchmark_name": "Humanity's Last Exam",
        "setting": "with-tools",
        "setting_name": "With tools",
        "score": 44.9,
        "unit": "%",
        "printed": "44.9",
        "reported_by": "moonshot",
        "run_by": null,
        "label": "K2 Thinking",
        "source": "https://www.kimi.ai/blog/kimi-k2-thinking",
        "source_title": "Introducing Kimi K2 Thinking",
        "source_date": "2025-11-06",
        "where": "'Full Evaluations' table, row 'Humanity's Last Exam (Text-only)', sub-row 'w/ tools [4]', column 'K2 Thinking'",
        "own": true,
        "primary": true,
        "parallel": false,
        "notes": "text-only subset; search, code and browsing tools; up to 120 steps; Hugging Face blocked; INT4 precision",
        "caveat": "",
        "human_verified": false,
        "human_verified_on": null,
        "check": "corroborated",
        "corroborated": true,
        "corroborated_on": "2026-10-19",
        "corroborated_by": [
          {
            "url": "https://huggingface.co/zai-org/GLM-4.7",
            "name": "GLM-4.7 card",
            "title": "zai-org/GLM-4.7 model card",
            "lab": "zai",
            "date": "2025-12-22",
            "printed": "44.9",
            "where": "README benchmark table, row 'HLE (w/ Tools)', column 'Kimi K2 Thinking'"
          }
        ],
        "source_capture": "https://frontierlog.ai/captures/4e2626572b0cef3e.png"
      },
      {
        "id": "kimi-k2-thinking:hle:with-tools:2",
        "model": "kimi-k2-thinking",
        "model_name": "Kimi K2 Thinking",
        "model_date": "2025-11-06",
        "event": "https://frontierlog.ai/event/2025-11-06-kimi-k2-thinking",
        "benchmark": "hle",
        "benchmark_name": "Humanity's Last Exam",
        "setting": "with-tools",
        "setting_name": "With tools",
        "score": 51.3,
        "unit": "%",
        "printed": "51.3",
        "reported_by": "moonshot",
        "run_by": null,
        "label": "K2 Thinking",
        "source": "https://www.kimi.ai/blog/kimi-k2-thinking",
        "source_title": "Introducing Kimi K2 Thinking",
        "source_date": "2025-11-06",
        "where": "Footnotes, 'For HLE (w/ tools) and the agentic-search benchmarks', item f (score without blocking Hugging Face)",
        "own": true,
        "primary": false,
        "parallel": false,
        "notes": "Hugging Face access not blocked, which may leak answers; text-only subset",
        "caveat": "",
        "human_verified": true,
        "human_verified_on": "2026-10-20",
        "check": "human",
        "corroborated": false,
        "corroborated_on": null,
        "corroborated_by": []
      },
      {
        "id": "kimi-k2-thinking:hle:with-tools:3",
        "model": "kimi-k2-thinking",
        "model_name": "Kimi K2 Thinking",
        "model_date": "2025-11-06",
        "event": "https://frontierlog.ai/event/2025-11-06-kimi-k2-thinking",
        "benchmark": "hle",
        "benchmark_name": "Humanity's Last Exam",
        "setting": "with-tools",
        "setting_name": "With tools",
        "score": 51.0,
        "unit": "%",
        "printed": "51.0",
        "reported_by": "moonshot",
        "run_by": null,
        "label": "K2 Thinking",
        "source": "https://www.kimi.ai/blog/kimi-k2-thinking",
        "source_title": "Introducing Kimi K2 Thinking",
        "source_date": "2025-11-06",
        "where": "'Full Evaluations' table, row 'Humanity's Last Exam (Text-only)', sub-row 'heavy [6]', column 'K2 Thinking'",
        "own": true,
        "primary": false,
        "parallel": true,
        "notes": "heavy mode: eight parallel trajectories aggregated into one answer; text-only subset",
        "caveat": "",
        "human_verified": true,
        "human_verified_on": "2026-10-20",
        "check": "human",
        "corroborated": false,
        "corroborated_on": null,
        "corroborated_by": [],
        "source_capture": "https://frontierlog.ai/captures/4e2626572b0cef3e.png"
      },
      {
        "id": "kimi-k2-thinking:gpqa-diamond:default:1",
        "model": "kimi-k2-thinking",
        "model_name": "Kimi K2 Thinking",
        "model_date": "2025-11-06",
        "event": "https://frontierlog.ai/event/2025-11-06-kimi-k2-thinking",
        "benchmark": "gpqa-diamond",
        "benchmark_name": "GPQA Diamond",
        "setting": "default",
        "setting_name": "",
        "score": 84.5,
        "unit": "%",
        "printed": "84.5",
        "reported_by": "moonshot",
        "run_by": null,
        "label": "K2 Thinking",
        "source": "https://www.kimi.ai/blog/kimi-k2-thinking",
        "source_title": "Introducing Kimi K2 Thinking",
        "source_date": "2025-11-06",
        "where": "'Full Evaluations' table, row 'GPQA-Diamond' (no tools), column 'K2 Thinking'",
        "own": true,
        "primary": true,
        "parallel": false,
        "notes": "96K thinking-token budget; INT4 precision",
        "caveat": "",
        "human_verified": false,
        "human_verified_on": null,
        "check": "corroborated",
        "corroborated": true,
        "corroborated_on": "2026-10-19",
        "corroborated_by": [
          {
            "url": "https://huggingface.co/zai-org/GLM-4.7",
            "name": "GLM-4.7 card",
            "title": "zai-org/GLM-4.7 model card",
            "lab": "zai",
            "date": "2025-12-22",
            "printed": "84.5",
            "where": "README benchmark table, row 'GPQA-Diamond', column 'Kimi K2 Thinking'"
          }
        ],
        "source_capture": "https://frontierlog.ai/captures/4e2626572b0cef3e.png"
      }
    ]
  }
}