{
  "api_version": "1",
  "updated": "2026-10-26",
  "latest_event": "2026-10-24",
  "rights": "© Frontier Log, all rights reserved.",
  "model": {
    "slug": "gpt-5-2",
    "name": "GPT-5.2",
    "lab": "openai",
    "date": "2025-12-11",
    "flagship": true,
    "event": "https://frontierlog.ai/event/2025-12-11-gpt-5-2",
    "releases": [],
    "variant_of": null,
    "made_by": "OpenAI",
    "url": "https://frontierlog.ai/model/gpt-5-2",
    "results": [
      {
        "id": "gpt-5-2:swe-bench-verified:default:1",
        "model": "gpt-5-2",
        "model_name": "GPT-5.2",
        "model_date": "2025-12-11",
        "event": "https://frontierlog.ai/event/2025-12-11-gpt-5-2",
        "benchmark": "swe-bench-verified",
        "benchmark_name": "SWE-bench Verified",
        "setting": "default",
        "setting_name": "",
        "score": 80.0,
        "unit": "%",
        "printed": "80.0%",
        "reported_by": "openai",
        "run_by": null,
        "label": "GPT‑5.2 Thinking",
        "source": "https://openai.com/index/introducing-gpt-5-2",
        "source_title": "Introducing GPT-5.2",
        "source_date": "2025-12-11",
        "where": "coding table, row 'SWE-bench Verified', column 'GPT-5.2 Thinking'",
        "own": true,
        "primary": true,
        "parallel": false,
        "notes": "",
        "caveat": "",
        "human_verified": true,
        "human_verified_on": "2026-10-16",
        "check": "human",
        "corroborated": true,
        "corroborated_on": "2026-10-19",
        "corroborated_by": [
          {
            "url": "https://www.anthropic.com/claude-opus-4-6-system-card",
            "name": "Claude Opus 4.6 card",
            "title": "Claude Opus 4.6 System Card",
            "lab": "anthropic",
            "date": "2026-02-05",
            "printed": "80.0%",
            "where": "Table 2.3.A (p.18), row SWE-bench Verified, column GPT-5.2"
          },
          {
            "url": "https://www.anthropic.com/news/claude-opus-4-6",
            "name": "Claude Opus 4.6",
            "title": "Claude Opus 4.6",
            "lab": "anthropic",
            "date": "2026-02-05",
            "printed": "80.0%",
            "where": "Benchmark table image, row 'Agentic coding SWE-bench Verified', column 'GPT-5.2 (all models)'"
          },
          {
            "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": "80",
            "where": "Coding bar chart image (10 panels); panel 'SWE-bench Verified'; value label on GPT-5.2 (thinking) bar"
          },
          {
            "url": "https://www.minimax.io/news/minimax-m25",
            "name": "MiniMax M2.5",
            "title": "MiniMax M2.5: Built for Real-World Productivity.",
            "lab": "minimax",
            "date": "2026-02-12",
            "printed": "80",
            "where": "Overview bar chart image (8 panels); panel 'SWE-Bench Verified'; value label on GPT-5.2 bar"
          },
          {
            "url": "https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-1-pro/",
            "name": "Gemini 3.1 Pro",
            "title": "Gemini 3.1 Pro: A smarter model for your most complex tasks",
            "lab": "google",
            "date": "2026-02-19",
            "printed": "80.0%",
            "where": "Benchmark table image; row 'SWE-Bench Verified, Single attempt'; column gpt-5-2"
          },
          {
            "url": "https://www.kimi.ai/blog/kimi-k2-5",
            "name": "Kimi K2.5",
            "title": "Kimi K2.5: Visual Agentic Intelligence",
            "lab": "moonshot",
            "date": "2026-01-27",
            "printed": "80.0",
            "where": "Appendix Benchmark table, row SWE-Bench Verified ; column GPT-5.2 (xhigh)"
          },
          {
            "url": "https://qwen.ai/blog?id=qwen3-max-thinking",
            "name": "Qwen3-Max-Thinking",
            "title": "Pushing Qwen3-Max-Thinking Beyond its Limits",
            "lab": "alibaba",
            "date": "2026-01-26",
            "printed": "80.0",
            "where": "Score chart image (score.png), panel SWE-bench Verified, value label on the GPT-5.2 bar"
          },
          {
            "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": "80.0",
            "where": "Benchmark table, row 'SWE-bench Verified', column 'GPT-5.2 (xhigh)'"
          }
        ],
        "source_capture": "https://frontierlog.ai/captures/da9899c90caed7cc.png"
      },
      {
        "id": "gpt-5-2:gpqa-diamond:default:1",
        "model": "gpt-5-2",
        "model_name": "GPT-5.2",
        "model_date": "2025-12-11",
        "event": "https://frontierlog.ai/event/2025-12-11-gpt-5-2",
        "benchmark": "gpqa-diamond",
        "benchmark_name": "GPQA Diamond",
        "setting": "default",
        "setting_name": "",
        "score": 92.4,
        "unit": "%",
        "printed": "92.4%",
        "reported_by": "openai",
        "run_by": null,
        "label": "GPT‑5.2 Thinking",
        "source": "https://openai.com/index/introducing-gpt-5-2",
        "source_title": "Introducing GPT-5.2",
        "source_date": "2025-12-11",
        "where": "science and math table, row 'GPQA Diamond (no tools)', column 'GPT-5.2 Thinking'",
        "own": true,
        "primary": true,
        "parallel": false,
        "notes": "no tools",
        "caveat": "",
        "human_verified": true,
        "human_verified_on": "2026-10-16",
        "check": "human",
        "corroborated": true,
        "corroborated_on": "2026-10-19",
        "corroborated_by": [
          {
            "url": "https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-1-pro/",
            "name": "Gemini 3.1 Pro",
            "title": "Gemini 3.1 Pro: A smarter model for your most complex tasks",
            "lab": "google",
            "date": "2026-02-19",
            "printed": "92.4%",
            "where": "Benchmark table image; row 'GPQA Diamond, No tools'; column gpt-5-2"
          },
          {
            "url": "https://www.kimi.ai/blog/kimi-k2-5",
            "name": "Kimi K2.5",
            "title": "Kimi K2.5: Visual Agentic Intelligence",
            "lab": "moonshot",
            "date": "2026-01-27",
            "printed": "92.4",
            "where": "Appendix Benchmark table, row GPQA-Diamond (avg@8) ; column GPT-5.2 (xhigh)"
          },
          {
            "url": "https://openai.com/index/introducing-gpt-5-4",
            "name": "GPT-5.4",
            "title": "Introducing GPT-5.4",
            "lab": "openai",
            "date": "2026-03-05",
            "printed": "92.4%",
            "where": "table 6 \"GPQA Diamond\" row, \"GPT-5.2\" column"
          },
          {
            "url": "https://qwen.ai/blog?id=qwen3-max-thinking",
            "name": "Qwen3-Max-Thinking",
            "title": "Pushing Qwen3-Max-Thinking Beyond its Limits",
            "lab": "alibaba",
            "date": "2026-01-26",
            "printed": "92.4",
            "where": "Score chart image (score.png), panel GPQA Diamond, value label on the GPT-5.2 bar"
          },
          {
            "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": "92.4",
            "where": "Benchmark table, row 'GPQA-Diamond', column 'GPT-5.2 (xhigh)'"
          }
        ],
        "source_capture": "https://frontierlog.ai/captures/22817d0b2b0a2432.png"
      },
      {
        "id": "gpt-5-2:hle:no-tools:1",
        "model": "gpt-5-2",
        "model_name": "GPT-5.2",
        "model_date": "2025-12-11",
        "event": "https://frontierlog.ai/event/2025-12-11-gpt-5-2",
        "benchmark": "hle",
        "benchmark_name": "Humanity's Last Exam",
        "setting": "no-tools",
        "setting_name": "No tools",
        "score": 34.5,
        "unit": "%",
        "printed": "34.5%",
        "reported_by": "openai",
        "run_by": null,
        "label": "GPT‑5.2 Thinking",
        "source": "https://openai.com/index/introducing-gpt-5-2",
        "source_title": "Introducing GPT-5.2",
        "source_date": "2025-12-11",
        "where": "science and math table, row 'HLE (no tools)', column 'GPT-5.2 Thinking'",
        "own": true,
        "primary": true,
        "parallel": false,
        "notes": "",
        "caveat": "",
        "human_verified": true,
        "human_verified_on": "2026-10-16",
        "check": "human",
        "corroborated": true,
        "corroborated_on": "2026-10-19",
        "corroborated_by": [
          {
            "url": "https://www.anthropic.com/news/claude-opus-4-6",
            "name": "Claude Opus 4.6",
            "title": "Claude Opus 4.6",
            "lab": "anthropic",
            "date": "2026-02-05",
            "printed": "34.5",
            "where": "Bar chart 'Multidisciplinary reasoning, Humanity's Last Exam', bar 'GPT-5.2', lighter 'Without tools' bar"
          },
          {
            "url": "https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-1-pro/",
            "name": "Gemini 3.1 Pro",
            "title": "Gemini 3.1 Pro: A smarter model for your most complex tasks",
            "lab": "google",
            "date": "2026-02-19",
            "printed": "34.5%",
            "where": "Benchmark table image; row 'Humanity's Last Exam, No tools'; column gpt-5-2"
          },
          {
            "url": "https://openai.com/index/introducing-gpt-5-4",
            "name": "GPT-5.4",
            "title": "Introducing GPT-5.4",
            "lab": "openai",
            "date": "2026-03-05",
            "printed": "34.5%",
            "where": "table 6 \"Humanity's Last Exam (no tools)\" row, \"GPT-5.2\" column"
          }
        ],
        "source_capture": "https://frontierlog.ai/captures/22817d0b2b0a2432.png"
      },
      {
        "id": "gpt-5-2:arc-agi-2:default:1",
        "model": "gpt-5-2",
        "model_name": "GPT-5.2",
        "model_date": "2025-12-11",
        "event": "https://frontierlog.ai/event/2025-12-11-gpt-5-2",
        "benchmark": "arc-agi-2",
        "benchmark_name": "ARC-AGI-2",
        "setting": "default",
        "setting_name": "",
        "score": 52.9,
        "unit": "%",
        "printed": "52.9%",
        "reported_by": "openai",
        "run_by": null,
        "label": "GPT‑5.2 Thinking",
        "source": "https://openai.com/index/introducing-gpt-5-2",
        "source_title": "Introducing GPT-5.2",
        "source_date": "2025-12-11",
        "where": "abstract reasoning table, row 'ARC-AGI-2 (Verified)', column 'GPT-5.2 Thinking'",
        "own": true,
        "primary": true,
        "parallel": false,
        "notes": "ARC Prize Verified",
        "caveat": "",
        "human_verified": true,
        "human_verified_on": "2026-10-16",
        "check": "human",
        "corroborated": true,
        "corroborated_on": "2026-10-19",
        "corroborated_by": [
          {
            "url": "https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-1-pro/",
            "name": "Gemini 3.1 Pro",
            "title": "Gemini 3.1 Pro: A smarter model for your most complex tasks",
            "lab": "google",
            "date": "2026-02-19",
            "printed": "52.9%",
            "where": "Benchmark table image; row 'ARC-AGI-2, ARC Prize Verified'; column gpt-5-2"
          },
          {
            "url": "https://openai.com/index/introducing-gpt-5-4",
            "name": "GPT-5.4",
            "title": "Introducing GPT-5.4",
            "lab": "openai",
            "date": "2026-03-05",
            "printed": "52.9%",
            "where": "table 8 \"ARC-AGI-2 (Verified)\" row, \"GPT-5.2\" column"
          }
        ],
        "source_capture": "https://frontierlog.ai/captures/5c5b6d32d6479128.png"
      },
      {
        "id": "gpt-5-2:terminal-bench-2-0:default:1",
        "model": "gpt-5-2",
        "model_name": "GPT-5.2",
        "model_date": "2025-12-11",
        "event": "https://frontierlog.ai/event/2025-12-11-gpt-5-2",
        "benchmark": "terminal-bench-2-0",
        "benchmark_name": "Terminal-Bench 2.0",
        "setting": "default",
        "setting_name": "",
        "score": 62.2,
        "unit": "%",
        "printed": "62.2%",
        "reported_by": "openai",
        "run_by": null,
        "label": "GPT‑5.2",
        "source": "https://openai.com/index/introducing-gpt-5-4",
        "source_title": "Introducing GPT-5.4",
        "source_date": "2026-03-05",
        "where": "coding table, row 'Terminal-Bench 2.0', column 'GPT‑5.2'",
        "own": false,
        "primary": true,
        "parallel": false,
        "notes": "harness not stated",
        "caveat": "",
        "human_verified": true,
        "human_verified_on": "2026-10-16",
        "check": "human",
        "corroborated": true,
        "corroborated_on": "2026-10-19",
        "corroborated_by": [
          {
            "url": "https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-1-pro/",
            "name": "Gemini 3.1 Pro",
            "title": "Gemini 3.1 Pro: A smarter model for your most complex tasks",
            "lab": "google",
            "date": "2026-02-19",
            "printed": "62.2%",
            "where": "Benchmark table image; row 'Terminal-Bench 2.0, Other best self-reported harness (Codex)'; column gpt-5-2"
          },
          {
            "url": "https://openai.com/index/introducing-gpt-5-3-codex",
            "name": "GPT-5.3-Codex",
            "title": "Introducing GPT-5.3-Codex",
            "lab": "openai",
            "date": "2026-02-05",
            "printed": "62.2%",
            "where": "table 0 \"Terminal-Bench 2.0\" row, \"GPT-5.2 (xhigh)\" column"
          }
        ],
        "source_capture": "https://frontierlog.ai/captures/9941c2d6c8891638.png"
      },
      {
        "id": "gpt-5-2:hle:with-tools:1",
        "model": "gpt-5-2",
        "model_name": "GPT-5.2",
        "model_date": "2025-12-11",
        "event": "https://frontierlog.ai/event/2025-12-11-gpt-5-2",
        "benchmark": "hle",
        "benchmark_name": "Humanity's Last Exam",
        "setting": "with-tools",
        "setting_name": "With tools",
        "score": 45.5,
        "unit": "%",
        "printed": "45.5%",
        "reported_by": "openai",
        "run_by": null,
        "label": "GPT-5.2 Thinking",
        "source": "https://openai.com/index/introducing-gpt-5-2",
        "source_title": "Introducing GPT-5.2",
        "source_date": "2025-12-11",
        "where": "science and math eval table, column 'GPT-5.2 Thinking', row 'HLE (w/ search, Python)'",
        "own": true,
        "primary": true,
        "parallel": false,
        "notes": "search and Python",
        "caveat": "",
        "human_verified": true,
        "human_verified_on": "2026-10-17",
        "check": "human",
        "corroborated": true,
        "corroborated_on": "2026-10-19",
        "corroborated_by": [
          {
            "url": "https://www.anthropic.com/claude-opus-4-6-system-card",
            "name": "Claude Opus 4.6 card",
            "title": "Claude Opus 4.6 System Card",
            "lab": "anthropic",
            "date": "2026-02-05",
            "printed": "45.5%",
            "where": "Figure 2.21.2.A (p.42), right panel With Tools, bar GPT-5.2 Thinking value label"
          },
          {
            "url": "https://www.anthropic.com/news/claude-opus-4-6",
            "name": "Claude Opus 4.6",
            "title": "Claude Opus 4.6",
            "lab": "anthropic",
            "date": "2026-02-05",
            "printed": "45.5",
            "where": "Bar chart 'Multidisciplinary reasoning, Humanity's Last Exam', bar 'GPT-5.2', darker 'With tools' bar"
          },
          {
            "url": "https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-1-pro/",
            "name": "Gemini 3.1 Pro",
            "title": "Gemini 3.1 Pro: A smarter model for your most complex tasks",
            "lab": "google",
            "date": "2026-02-19",
            "printed": "45.5%",
            "where": "Benchmark table image; row 'Humanity's Last Exam, Search (blocklist) + Code'; column gpt-5-2"
          },
          {
            "url": "https://www.kimi.ai/blog/kimi-k2-5",
            "name": "Kimi K2.5",
            "title": "Kimi K2.5: Visual Agentic Intelligence",
            "lab": "moonshot",
            "date": "2026-01-27",
            "printed": "45.5",
            "where": "Appendix Benchmark table, row HLE-Full w/ tools ; column GPT-5.2 (xhigh)"
          },
          {
            "url": "https://openai.com/index/introducing-gpt-5-4",
            "name": "GPT-5.4",
            "title": "Introducing GPT-5.4",
            "lab": "openai",
            "date": "2026-03-05",
            "printed": "45.5%",
            "where": "table 6 \"Humanity's Last Exam (with tools)\" row, \"GPT-5.2\" column"
          },
          {
            "url": "https://qwen.ai/blog?id=qwen3-max-thinking",
            "name": "Qwen3-Max-Thinking",
            "title": "Pushing Qwen3-Max-Thinking Beyond its Limits",
            "lab": "alibaba",
            "date": "2026-01-26",
            "printed": "45.5",
            "where": "Score chart image (score.png), panel Humanity's Last Exam (with Search), value label on the GPT-5.2 bar"
          },
          {
            "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": "45.5",
            "where": "Benchmark table, row 'Humanity's Last Exam w/ Tools (* = full set for Opus 4.5, Gemini, GPT-5.2)', column 'GPT-5.2 (xhigh)'"
          }
        ],
        "source_capture": "https://frontierlog.ai/captures/22817d0b2b0a2432.png"
      },
      {
        "id": "gpt-5-2:browsecomp:default:1",
        "model": "gpt-5-2",
        "model_name": "GPT-5.2",
        "model_date": "2025-12-11",
        "event": "https://frontierlog.ai/event/2025-12-11-gpt-5-2",
        "benchmark": "browsecomp",
        "benchmark_name": "BrowseComp",
        "setting": "default",
        "setting_name": "",
        "score": 65.8,
        "unit": "%",
        "printed": "65.8%",
        "reported_by": "openai",
        "run_by": null,
        "label": "GPT-5.2 Thinking",
        "source": "https://openai.com/index/introducing-gpt-5-2",
        "source_title": "Introducing GPT-5.2",
        "source_date": "2025-12-11",
        "where": "agentic tool-use table (columns GPT-5.2 Thinking / GPT-5.2 Pro / GPT-5.1 Thinking), row 'BrowseComp', column 'GPT-5.2 Thinking'",
        "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.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": "65.8",
            "where": "Tool-use and intelligence table image; row 'BrowseComp'; column 'GPT-5.2 (thinking)'"
          },
          {
            "url": "https://www.minimax.io/news/minimax-m25",
            "name": "MiniMax M2.5",
            "title": "MiniMax M2.5: Built for Real-World Productivity.",
            "lab": "minimax",
            "date": "2026-02-12",
            "printed": "65.8",
            "where": "Overview bar chart image (8 panels); panel 'BrowseComp (w/ctx)'; value label on GPT-5.2 bar"
          },
          {
            "url": "https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-1-pro/",
            "name": "Gemini 3.1 Pro",
            "title": "Gemini 3.1 Pro: A smarter model for your most complex tasks",
            "lab": "google",
            "date": "2026-02-19",
            "printed": "65.8%",
            "where": "Benchmark table image; row 'BrowseComp, Search + Python + Browse'; column gpt-5-2"
          },
          {
            "url": "https://www.kimi.ai/blog/kimi-k2-5",
            "name": "Kimi K2.5",
            "title": "Kimi K2.5: Visual Agentic Intelligence",
            "lab": "moonshot",
            "date": "2026-01-27",
            "printed": "65.8",
            "where": "Appendix Benchmark table, row BrowseComp (w/ctx mgm) ; column GPT-5.2 (xhigh)"
          },
          {
            "url": "https://openai.com/index/introducing-gpt-5-4",
            "name": "GPT-5.4",
            "title": "Introducing GPT-5.4",
            "lab": "openai",
            "date": "2026-03-05",
            "printed": "65.8%",
            "where": "table 0 \"BrowseComp\" row, \"GPT-5.2\" column"
          },
          {
            "url": "https://qwen.ai/blog?id=qwen3.5",
            "name": "Qwen3.5",
            "title": "Qwen3.5: Towards Native Multimodal Agents",
            "lab": "alibaba",
            "date": "2026-02-16",
            "printed": "65.8",
            "where": "Score chart image (qwen3.5_397b_a17b_score.png), panel BrowseComp, value label on the GPT-5.2 bar"
          },
          {
            "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": "65.8",
            "where": "Benchmark table, row 'BrowseComp w/ Context Manage', column 'GPT-5.2 (xhigh)'"
          }
        ],
        "source_capture": "https://frontierlog.ai/captures/a408379f5cd396d1.png"
      },
      {
        "id": "gpt-5-2:swe-bench-pro:default:1",
        "model": "gpt-5-2",
        "model_name": "GPT-5.2",
        "model_date": "2025-12-11",
        "event": "https://frontierlog.ai/event/2025-12-11-gpt-5-2",
        "benchmark": "swe-bench-pro",
        "benchmark_name": "SWE-bench Pro",
        "setting": "default",
        "setting_name": "",
        "score": 55.6,
        "unit": "%",
        "printed": "55.6%",
        "reported_by": "openai",
        "run_by": null,
        "label": "GPT-5.2 Thinking",
        "source": "https://openai.com/index/introducing-gpt-5-2",
        "source_title": "Introducing GPT-5.2",
        "source_date": "2025-12-11",
        "where": "coding table (columns GPT-5.2 Thinking, GPT-5.2 Pro, GPT-5.1 Thinking), row 'SWE-Bench Pro, Public', column 'GPT-5.2 Thinking'",
        "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.minimax.io/news/minimax-m25",
            "name": "MiniMax M2.5",
            "title": "MiniMax M2.5: Built for Real-World Productivity.",
            "lab": "minimax",
            "date": "2026-02-12",
            "printed": "55.6",
            "where": "Overview bar chart image (8 panels); panel 'SWE-Bench Pro'; value label on GPT-5.2 bar"
          },
          {
            "url": "https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-1-pro/",
            "name": "Gemini 3.1 Pro",
            "title": "Gemini 3.1 Pro: A smarter model for your most complex tasks",
            "lab": "google",
            "date": "2026-02-19",
            "printed": "55.6%",
            "where": "Benchmark table image; row 'SWE-Bench Pro (Public), Single attempt'; column gpt-5-2"
          },
          {
            "url": "https://www.kimi.ai/blog/kimi-k2-5",
            "name": "Kimi K2.5",
            "title": "Kimi K2.5: Visual Agentic Intelligence",
            "lab": "moonshot",
            "date": "2026-01-27",
            "printed": "55.6",
            "where": "Appendix Benchmark table, row SWE-Bench Pro ; column GPT-5.2 (xhigh)"
          },
          {
            "url": "https://openai.com/index/introducing-gpt-5-4",
            "name": "GPT-5.4",
            "title": "Introducing GPT-5.4",
            "lab": "openai",
            "date": "2026-03-05",
            "printed": "55.6%",
            "where": "table 0 \"SWE-Bench Pro (Public)\" row, \"GPT-5.2\" column"
          },
          {
            "url": "https://openai.com/index/introducing-gpt-5-3-codex",
            "name": "GPT-5.3-Codex",
            "title": "Introducing GPT-5.3-Codex",
            "lab": "openai",
            "date": "2026-02-05",
            "printed": "55.6%",
            "where": "table 0 \"SWE-Bench Pro (Public)\" row, \"GPT-5.2 (xhigh)\" column"
          }
        ],
        "source_capture": "https://frontierlog.ai/captures/da9899c90caed7cc.png"
      },
      {
        "id": "gpt-5-2:aime-2025:no-tools:1",
        "model": "gpt-5-2",
        "model_name": "GPT-5.2",
        "model_date": "2025-12-11",
        "event": "https://frontierlog.ai/event/2025-12-11-gpt-5-2",
        "benchmark": "aime-2025",
        "benchmark_name": "AIME 2025",
        "setting": "no-tools",
        "setting_name": "No tools",
        "score": 100.0,
        "unit": "%",
        "printed": "100.0%",
        "reported_by": "openai",
        "run_by": null,
        "label": "GPT‑5.2 Thinking",
        "source": "https://openai.com/index/introducing-gpt-5-2",
        "source_title": "Introducing GPT-5.2",
        "source_date": "2025-12-11",
        "where": "headline eval table (HTML, first table), row 'AIME 2025 (no tools) / Competition math', column 'GPT-5.2 Thinking'",
        "own": true,
        "primary": true,
        "parallel": false,
        "notes": "",
        "caveat": "",
        "human_verified": true,
        "human_verified_on": "2026-10-18",
        "check": "human",
        "corroborated": false,
        "corroborated_on": null,
        "corroborated_by": [],
        "source_capture": "https://frontierlog.ai/captures/22817d0b2b0a2432.png"
      },
      {
        "id": "gpt-5-2:arc-agi-1:default:1",
        "model": "gpt-5-2",
        "model_name": "GPT-5.2",
        "model_date": "2025-12-11",
        "event": "https://frontierlog.ai/event/2025-12-11-gpt-5-2",
        "benchmark": "arc-agi-1",
        "benchmark_name": "ARC-AGI-1",
        "setting": "default",
        "setting_name": "",
        "score": 86.2,
        "unit": "%",
        "printed": "86.2%",
        "reported_by": "openai",
        "run_by": null,
        "label": "GPT‑5.2 Thinking",
        "source": "https://openai.com/index/introducing-gpt-5-2",
        "source_title": "Introducing GPT-5.2",
        "source_date": "2025-12-11",
        "where": "headline eval table (HTML, first table), row 'ARC-AGI-1 (Verified) / Abstract reasoning', column 'GPT‑5.2 Thinking'",
        "own": true,
        "primary": true,
        "parallel": false,
        "notes": "ARC Prize Verified",
        "caveat": "",
        "human_verified": true,
        "human_verified_on": "2026-10-18",
        "check": "human",
        "corroborated": true,
        "corroborated_on": "2026-10-19",
        "corroborated_by": [
          {
            "url": "https://openai.com/index/introducing-gpt-5-4",
            "name": "GPT-5.4",
            "title": "Introducing GPT-5.4",
            "lab": "openai",
            "date": "2026-03-05",
            "printed": "86.2%",
            "where": "table 8 \"ARC-AGI-1 (Verified)\" row, \"GPT-5.2\" column"
          }
        ],
        "source_capture": "https://frontierlog.ai/captures/5c5b6d32d6479128.png"
      },
      {
        "id": "gpt-5-2:hmmt-2025:no-tools:1",
        "model": "gpt-5-2",
        "model_name": "GPT-5.2",
        "model_date": "2025-12-11",
        "event": "https://frontierlog.ai/event/2025-12-11-gpt-5-2",
        "benchmark": "hmmt-2025",
        "benchmark_name": "HMMT February 2025",
        "setting": "no-tools",
        "setting_name": "No tools",
        "score": 99.4,
        "unit": "%",
        "printed": "99.4%",
        "reported_by": "openai",
        "run_by": null,
        "label": "GPT-5.2 Thinking",
        "source": "https://openai.com/index/introducing-gpt-5-2",
        "source_title": "Introducing GPT-5.2",
        "source_date": "2025-12-11",
        "where": "benchmark table (academic evals), row 'HMMT, Feb 2025 (no tools)', column 'GPT-5.2 Thinking'",
        "own": true,
        "primary": true,
        "parallel": false,
        "notes": "maximum reasoning effort",
        "caveat": "",
        "human_verified": true,
        "human_verified_on": "2026-10-18",
        "check": "human",
        "corroborated": true,
        "corroborated_on": "2026-10-19",
        "corroborated_by": [
          {
            "url": "https://www.kimi.ai/blog/kimi-k2-5",
            "name": "Kimi K2.5",
            "title": "Kimi K2.5: Visual Agentic Intelligence",
            "lab": "moonshot",
            "date": "2026-01-27",
            "printed": "99.4",
            "where": "Appendix Benchmark table, row HMMT 2025 (Feb) (avg@32) ; column GPT-5.2 (xhigh)"
          }
        ],
        "source_capture": "https://frontierlog.ai/captures/22817d0b2b0a2432.png"
      }
    ]
  }
}