{
  "api_version": "1",
  "updated": "2026-10-26",
  "latest_event": "2026-10-24",
  "rights": "© Frontier Log, all rights reserved.",
  "model": {
    "slug": "kimi-k2-5",
    "name": "Kimi K2.5",
    "lab": "moonshot",
    "date": "2026-01-27",
    "flagship": false,
    "event": "https://frontierlog.ai/event/2026-01-27-kimi-k2-5",
    "releases": [],
    "variant_of": null,
    "made_by": "Moonshot AI",
    "url": "https://frontierlog.ai/model/kimi-k2-5",
    "results": [
      {
        "id": "kimi-k2-5:swe-bench-verified:default:1",
        "model": "kimi-k2-5",
        "model_name": "Kimi K2.5",
        "model_date": "2026-01-27",
        "event": "https://frontierlog.ai/event/2026-01-27-kimi-k2-5",
        "benchmark": "swe-bench-verified",
        "benchmark_name": "SWE-bench Verified",
        "setting": "default",
        "setting_name": "",
        "score": 76.8,
        "unit": "%",
        "printed": "76.8",
        "reported_by": "moonshot",
        "run_by": null,
        "label": "Kimi K2.5",
        "source": "https://www.kimi.ai/blog/kimi-k2-5",
        "source_title": "Kimi K2.5: Visual Agentic Intelligence",
        "source_date": "2026-01-27",
        "where": "benchmark chart 'Coding', 'SWE-Bench Verified', bar 'Kimi K2.5'",
        "own": true,
        "primary": true,
        "parallel": false,
        "notes": "",
        "caveat": "",
        "human_verified": true,
        "human_verified_on": "2026-10-17",
        "check": "human",
        "corroborated": true,
        "corroborated_on": "2026-10-19",
        "corroborated_by": [
          {
            "url": "https://www.kimi.ai/blog/kimi-k2-6",
            "name": "Kimi K2.6",
            "title": "Kimi K2.6: Advancing Open-Source Coding",
            "lab": "moonshot",
            "date": "2026-04-20",
            "printed": "76.8",
            "where": "Benchmark Table (appendix), row SWE-Bench Verified ; column Kimi K2.5"
          },
          {
            "url": "https://qwen.ai/blog?id=qwen3.6",
            "name": "Qwen3.6-Plus",
            "title": "Qwen3.6-Plus: Towards Real World Agents",
            "lab": "alibaba",
            "date": "2026-04-02",
            "printed": "76.8",
            "where": "Score chart image (qwen3.6_plus_score.png), panel SWE-bench Verified, value label on the Kimi K2.5 bar (bars identified by legend colour and logo)"
          },
          {
            "url": "https://mistral.ai/news/vibe-remote-agents-mistral-medium-3-5/",
            "name": "Mistral Medium 3.5",
            "title": "Remote agents in Vibe. Powered by Mistral Medium 3.5.",
            "lab": "mistral",
            "date": "2026-05-22",
            "printed": "76.8",
            "where": "Chart 'Agentic Benchmarks vs competing models', group 'SWE-Bench verified***' (footnote: self-reported; max reasoning), bar 'Kimi K2.5 1T A32B'"
          },
          {
            "url": "https://thinkingmachines.ai/news/introducing-inkling/",
            "name": "Inkling",
            "title": "Inkling: Our Open-Weights Model",
            "lab": "thinking-machines",
            "date": "2026-07-15",
            "printed": "76.8%",
            "where": "'Benchmarking Inkling' main table, row 'SWEBench Verified* (Inkling bash-only harness; external models self-reported)', column 'Kimi K2.5'"
          },
          {
            "url": "https://z.ai/blog/glm-5",
            "name": "GLM-5",
            "title": "GLM-5: From Vibe Coding to Agentic Engineering",
            "lab": "zai",
            "date": "2026-02-12",
            "printed": "76.8",
            "where": "Benchmark table, row 'SWE-bench Verified', column 'Kimi K2.5 (Thinking)'"
          }
        ],
        "source_capture": "https://frontierlog.ai/captures/a8cc819ddd0da74b.png"
      },
      {
        "id": "kimi-k2-5:hle:no-tools:1",
        "model": "kimi-k2-5",
        "model_name": "Kimi K2.5",
        "model_date": "2026-01-27",
        "event": "https://frontierlog.ai/event/2026-01-27-kimi-k2-5",
        "benchmark": "hle",
        "benchmark_name": "Humanity's Last Exam",
        "setting": "no-tools",
        "setting_name": "No tools",
        "score": 30.1,
        "unit": "%",
        "printed": "30.1",
        "reported_by": "moonshot",
        "run_by": null,
        "label": "Kimi K2.5 (Thinking)",
        "source": "https://www.kimi.ai/blog/kimi-k2-5",
        "source_title": "Kimi K2.5: Visual Agentic Intelligence",
        "source_date": "2026-01-27",
        "where": "Appendix 'Benchmark table', row 'HLE-Full' (Reasoning & Knowledge), column 'Kimi K2.5 (Thinking)'",
        "own": true,
        "primary": true,
        "parallel": false,
        "notes": "full set, text and image (31.5 text-only); 96K-token completion budget",
        "caveat": "",
        "human_verified": true,
        "human_verified_on": "2026-10-17",
        "check": "human",
        "corroborated": true,
        "corroborated_on": "2026-10-19",
        "corroborated_by": [
          {
            "url": "https://www.kimi.ai/blog/kimi-k2-6",
            "name": "Kimi K2.6",
            "title": "Kimi K2.6: Advancing Open-Source Coding",
            "lab": "moonshot",
            "date": "2026-04-20",
            "printed": "30.1",
            "where": "Benchmark Table (appendix), row HLE-Full (Reasoning & Knowledge; no tools) ; column Kimi K2.5"
          }
        ],
        "source_capture": "https://frontierlog.ai/captures/a8cc819ddd0da74b.png"
      },
      {
        "id": "kimi-k2-5:browsecomp:default:1",
        "model": "kimi-k2-5",
        "model_name": "Kimi K2.5",
        "model_date": "2026-01-27",
        "event": "https://frontierlog.ai/event/2026-01-27-kimi-k2-5",
        "benchmark": "browsecomp",
        "benchmark_name": "BrowseComp",
        "setting": "default",
        "setting_name": "",
        "score": 74.9,
        "unit": "%",
        "printed": "74.9",
        "reported_by": "moonshot",
        "run_by": null,
        "label": "Kimi K2.5 (Thinking)",
        "source": "https://www.kimi.ai/blog/kimi-k2-5",
        "source_title": "Kimi K2.5: Visual Agentic Intelligence",
        "source_date": "2026-01-27",
        "where": "Appendix 'Benchmark table', row 'BrowseComp (w/ctx mgm)' (Agentic Search), column 'Kimi K2.5 (Thinking)'",
        "own": true,
        "primary": true,
        "parallel": false,
        "notes": "with context management, discard-all strategy",
        "caveat": "",
        "human_verified": true,
        "human_verified_on": "2026-10-17",
        "check": "human",
        "corroborated": true,
        "corroborated_on": "2026-10-19",
        "corroborated_by": [
          {
            "url": "https://www.kimi.ai/blog/kimi-k2-6",
            "name": "Kimi K2.6",
            "title": "Kimi K2.6: Advancing Open-Source Coding",
            "lab": "moonshot",
            "date": "2026-04-20",
            "printed": "74.9",
            "where": "Benchmark Table (appendix), row BrowseComp ; column Kimi K2.5"
          },
          {
            "url": "https://mistral.ai/news/vibe-remote-agents-mistral-medium-3-5/",
            "name": "Mistral Medium 3.5",
            "title": "Remote agents in Vibe. Powered by Mistral Medium 3.5.",
            "lab": "mistral",
            "date": "2026-05-22",
            "printed": "74.9",
            "where": "Chart 'Agentic Benchmarks vs competing models', group 'BrowseComp****' (footnote: self-reported, Mistral uses context management, discard-all at 100k), bar 'Kimi K2.5 1T A32B'"
          },
          {
            "url": "https://thinkingmachines.ai/news/introducing-inkling/",
            "name": "Inkling",
            "title": "Inkling: Our Open-Weights Model",
            "lab": "thinking-machines",
            "date": "2026-07-15",
            "printed": "74.9%",
            "where": "'Benchmarking Inkling' main table, row 'BrowseComp w/ ctx management', column 'Kimi K2.5'"
          },
          {
            "url": "https://z.ai/blog/glm-5",
            "name": "GLM-5",
            "title": "GLM-5: From Vibe Coding to Agentic Engineering",
            "lab": "zai",
            "date": "2026-02-12",
            "printed": "74.9",
            "where": "Benchmark table, row 'BrowseComp w/ Context Manage', column 'Kimi K2.5 (Thinking)'"
          },
          {
            "url": "https://z.ai/blog/glm-5.1",
            "name": "GLM-5.1",
            "title": "GLM-5.1: Towards Long-Horizon Tasks",
            "lab": "zai",
            "date": "2026-04-07",
            "printed": "74.9",
            "where": "Benchmark table, row 'BrowseComp w/ Context Manage', column 'Kimi K2.5'"
          }
        ],
        "source_capture": "https://frontierlog.ai/captures/a8cc819ddd0da74b.png"
      },
      {
        "id": "kimi-k2-5:browsecomp:default:2",
        "model": "kimi-k2-5",
        "model_name": "Kimi K2.5",
        "model_date": "2026-01-27",
        "event": "https://frontierlog.ai/event/2026-01-27-kimi-k2-5",
        "benchmark": "browsecomp",
        "benchmark_name": "BrowseComp",
        "setting": "default",
        "setting_name": "",
        "score": 60.6,
        "unit": "%",
        "printed": "60.6",
        "reported_by": "moonshot",
        "run_by": null,
        "label": "Kimi K2.5 (Thinking)",
        "source": "https://www.kimi.ai/blog/kimi-k2-5",
        "source_title": "Kimi K2.5: Visual Agentic Intelligence",
        "source_date": "2026-01-27",
        "where": "Appendix 'Benchmark table', first row 'BrowseComp' (no qualifier, Agentic Search), column 'Kimi K2.5 (Thinking)'",
        "own": true,
        "primary": false,
        "parallel": false,
        "notes": "no context management: tasks past the context length count as failed",
        "caveat": "",
        "human_verified": true,
        "human_verified_on": "2026-10-17",
        "check": "human",
        "corroborated": true,
        "corroborated_on": "2026-10-19",
        "corroborated_by": [
          {
            "url": "https://z.ai/blog/glm-5",
            "name": "GLM-5",
            "title": "GLM-5: From Vibe Coding to Agentic Engineering",
            "lab": "zai",
            "date": "2026-02-12",
            "printed": "60.6",
            "where": "Benchmark table, row 'BrowseComp (no context management)', column 'Kimi K2.5 (Thinking)'"
          },
          {
            "url": "https://z.ai/blog/glm-5.1",
            "name": "GLM-5.1",
            "title": "GLM-5.1: Towards Long-Horizon Tasks",
            "lab": "zai",
            "date": "2026-04-07",
            "printed": "60.6",
            "where": "Benchmark table, row 'BrowseComp (no context management)', column 'Kimi K2.5'"
          }
        ],
        "source_capture": "https://frontierlog.ai/captures/a8cc819ddd0da74b.png"
      },
      {
        "id": "kimi-k2-5:swe-bench-pro:default:1",
        "model": "kimi-k2-5",
        "model_name": "Kimi K2.5",
        "model_date": "2026-01-27",
        "event": "https://frontierlog.ai/event/2026-01-27-kimi-k2-5",
        "benchmark": "swe-bench-pro",
        "benchmark_name": "SWE-bench Pro",
        "setting": "default",
        "setting_name": "",
        "score": 50.7,
        "unit": "%",
        "printed": "50.7",
        "reported_by": "moonshot",
        "run_by": null,
        "label": "Kimi K2.5 (Thinking)",
        "source": "https://www.kimi.ai/blog/kimi-k2-5",
        "source_title": "Kimi K2.5: Visual Agentic Intelligence",
        "source_date": "2026-01-27",
        "where": "Appendix 'Benchmark table', row 'SWE-Bench Pro' (Coding), column 'Kimi K2.5 (Thinking)' (read from the page's embedded benchmark JSON and the rendered table)",
        "own": true,
        "primary": true,
        "parallel": false,
        "notes": "in-house minimal-tool harness; non-thinking mode; average of 5 runs",
        "caveat": "The source doesn't say whether this is the public or the commercial set",
        "human_verified": true,
        "human_verified_on": "2026-10-17",
        "check": "human",
        "corroborated": true,
        "corroborated_on": "2026-10-19",
        "corroborated_by": [
          {
            "url": "https://www.kimi.ai/blog/kimi-k2-6",
            "name": "Kimi K2.6",
            "title": "Kimi K2.6: Advancing Open-Source Coding",
            "lab": "moonshot",
            "date": "2026-04-20",
            "printed": "50.7",
            "where": "Benchmark Table (appendix), row SWE-Bench Pro ; column Kimi K2.5"
          },
          {
            "url": "https://thinkingmachines.ai/news/introducing-inkling/",
            "name": "Inkling",
            "title": "Inkling: Our Open-Weights Model",
            "lab": "thinking-machines",
            "date": "2026-07-15",
            "printed": "50.7%",
            "where": "'Benchmarking Inkling' main table, row 'SWEBench Pro Public', column 'Kimi K2.5'"
          }
        ],
        "source_capture": "https://frontierlog.ai/captures/a8cc819ddd0da74b.png"
      },
      {
        "id": "kimi-k2-5:hle:with-tools:1",
        "model": "kimi-k2-5",
        "model_name": "Kimi K2.5",
        "model_date": "2026-01-27",
        "event": "https://frontierlog.ai/event/2026-01-27-kimi-k2-5",
        "benchmark": "hle",
        "benchmark_name": "Humanity's Last Exam",
        "setting": "with-tools",
        "setting_name": "With tools",
        "score": 50.2,
        "unit": "%",
        "printed": "50.2",
        "reported_by": "moonshot",
        "run_by": null,
        "label": "Kimi K2.5 (Thinking)",
        "source": "https://www.kimi.ai/blog/kimi-k2-5",
        "source_title": "Kimi K2.5: Visual Agentic Intelligence",
        "source_date": "2026-01-27",
        "where": "Appendix 'Benchmark table', row 'HLE-Full w/ tools' (Reasoning & Knowledge), column 'Kimi K2.5 (Thinking)'",
        "own": true,
        "primary": true,
        "parallel": false,
        "notes": "full set, text and image (51.8 text-only); search, code and browsing tools; Hugging Face blocked",
        "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/a8cc819ddd0da74b.png"
      },
      {
        "id": "kimi-k2-5:gpqa-diamond:default:1",
        "model": "kimi-k2-5",
        "model_name": "Kimi K2.5",
        "model_date": "2026-01-27",
        "event": "https://frontierlog.ai/event/2026-01-27-kimi-k2-5",
        "benchmark": "gpqa-diamond",
        "benchmark_name": "GPQA Diamond",
        "setting": "default",
        "setting_name": "",
        "score": 87.6,
        "unit": "%",
        "printed": "87.6",
        "reported_by": "moonshot",
        "run_by": null,
        "label": "Kimi K2.5 (Thinking)",
        "source": "https://www.kimi.ai/blog/kimi-k2-5",
        "source_title": "Kimi K2.5: Visual Agentic Intelligence",
        "source_date": "2026-01-27",
        "where": "Appendix 'Benchmark table', row 'GPQA-Diamond' (Reasoning & Knowledge), column 'Kimi K2.5 (Thinking)'",
        "own": true,
        "primary": true,
        "parallel": false,
        "notes": "thinking mode; 96K-token completion budget; average of 8 runs",
        "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/a8cc819ddd0da74b.png"
      }
    ]
  }
}