{
  "api_version": "1",
  "updated": "2026-10-26",
  "latest_event": "2026-10-24",
  "rights": "© Frontier Log, all rights reserved.",
  "model": {
    "slug": "gpt-5",
    "name": "GPT-5",
    "lab": "openai",
    "date": "2025-08-07",
    "flagship": true,
    "event": "https://frontierlog.ai/event/2025-08-07-gpt-5",
    "releases": [],
    "variant_of": null,
    "made_by": "OpenAI",
    "url": "https://frontierlog.ai/model/gpt-5",
    "results": [
      {
        "id": "gpt-5:swe-bench-verified:default:1",
        "model": "gpt-5",
        "model_name": "GPT-5",
        "model_date": "2025-08-07",
        "event": "https://frontierlog.ai/event/2025-08-07-gpt-5",
        "benchmark": "swe-bench-verified",
        "benchmark_name": "SWE-bench Verified",
        "setting": "default",
        "setting_name": "",
        "score": 74.9,
        "unit": "%",
        "printed": "74.9%",
        "reported_by": "openai",
        "run_by": null,
        "label": "GPT-5",
        "source": "https://openai.com/index/introducing-gpt-5",
        "source_title": "Introducing GPT-5",
        "source_date": "2025-08-07",
        "where": "text, opening summary of results",
        "own": true,
        "primary": true,
        "parallel": false,
        "notes": "477 of the 500 tasks",
        "caveat": "",
        "human_verified": true,
        "human_verified_on": "2026-10-16",
        "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": "74.9",
            "where": "Bar chart image; group 'SWE Verified (Resolved)'; value label on GPT-5-High bar"
          },
          {
            "url": "https://www.minimax.io/blog/minimax-m2-en-1748600000",
            "name": "MiniMax M2",
            "title": "MiniMax M2 & Agent: Ingenious in Simplicity",
            "lab": "minimax",
            "date": "2025-10-27",
            "printed": "74.9",
            "where": "Bar chart image (8 panels); panel 'SWE-bench Verified'; value label on GPT-5 (thinking) bar (legend order)"
          },
          {
            "url": "https://www.kimi.ai/blog/kimi-k2-thinking",
            "name": "Kimi K2 Thinking",
            "title": "Introducing Kimi K2 Thinking",
            "lab": "moonshot",
            "date": "2025-11-06",
            "printed": "74.9",
            "where": "Full Evaluations table, row SWE-bench Verified, w/ tools ; column GPT-5 (High)"
          },
          {
            "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": "74.9",
            "where": "Benchmark table, row 'SWE-bench Verified', column 'GPT-5 High'"
          }
        ]
      },
      {
        "id": "gpt-5:aime-2025:no-tools:1",
        "model": "gpt-5",
        "model_name": "GPT-5",
        "model_date": "2025-08-07",
        "event": "https://frontierlog.ai/event/2025-08-07-gpt-5",
        "benchmark": "aime-2025",
        "benchmark_name": "AIME 2025",
        "setting": "no-tools",
        "setting_name": "No tools",
        "score": 94.6,
        "unit": "%",
        "printed": "94.6%",
        "reported_by": "openai",
        "run_by": null,
        "label": "GPT‑5",
        "source": "https://openai.com/index/introducing-gpt-5",
        "source_title": "Introducing GPT-5",
        "source_date": "2025-08-07",
        "where": "'Evaluations' section, first paragraph (state of the art across math)",
        "own": true,
        "primary": true,
        "parallel": false,
        "notes": "with thinking; pass@1",
        "caveat": "",
        "human_verified": true,
        "human_verified_on": "2026-10-18",
        "check": "human",
        "corroborated": true,
        "corroborated_on": "2026-10-19",
        "corroborated_by": [
          {
            "url": "https://www.anthropic.com/news/claude-sonnet-4-5",
            "name": "Claude Agent SDK",
            "title": "Introducing Claude Sonnet 4.5",
            "lab": "anthropic",
            "date": "2025-09-29",
            "printed": "94.6%",
            "where": "Benchmark table image, row 'High school math competition AIME 2025', column 'GPT-5', lower value '(no tools)'"
          },
          {
            "url": "https://api-docs.deepseek.com/news/news251201/",
            "name": "DeepSeek-V3.2",
            "title": "DeepSeek-V3.2 Release",
            "lab": "deepseek",
            "date": "2025-12-01",
            "printed": "94.6",
            "where": "Bar chart image; group 'AIME 2025 (Pass@1)'; value label on GPT-5-High bar"
          },
          {
            "url": "https://www.kimi.ai/blog/kimi-k2-thinking",
            "name": "Kimi K2 Thinking",
            "title": "Introducing Kimi K2 Thinking",
            "lab": "moonshot",
            "date": "2025-11-06",
            "printed": "94.6",
            "where": "Full Evaluations table, row AIME 2025, no tools (avg@32) ; column GPT-5 (High)"
          },
          {
            "url": "https://openai.com/index/gpt-5-1-for-developers",
            "name": "GPT-5.1 in the API",
            "title": "Introducing GPT-5.1 for developers",
            "lab": "openai",
            "date": "2025-11-13",
            "printed": "94.6%",
            "where": "table 0 \"AIME 2025(no tools)\" row, \"GPT-5 (high)\" column"
          }
        ]
      },
      {
        "id": "gpt-5:aime-2025:no-tools:2",
        "model": "gpt-5",
        "model_name": "GPT-5",
        "model_date": "2025-08-07",
        "event": "https://frontierlog.ai/event/2025-08-07-gpt-5",
        "benchmark": "aime-2025",
        "benchmark_name": "AIME 2025",
        "setting": "no-tools",
        "setting_name": "No tools",
        "score": 61.9,
        "unit": "%",
        "printed": "61.9%",
        "reported_by": "openai",
        "run_by": null,
        "label": "GPT-5 (no tools), Without thinking",
        "source": "https://openai.com/index/introducing-gpt-5",
        "source_title": "Introducing GPT-5",
        "source_date": "2025-08-07",
        "where": "'Evaluations' section, bar chart 'AIME 2025 / Competition math', 4th bar 'GPT-5 (no tools)', lower (light) segment",
        "own": true,
        "primary": false,
        "parallel": false,
        "notes": "without thinking; pass@1",
        "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/a16f4aa39159111d.png"
      },
      {
        "id": "gpt-5:aime-2025:with-tools:1",
        "model": "gpt-5",
        "model_name": "GPT-5",
        "model_date": "2025-08-07",
        "event": "https://frontierlog.ai/event/2025-08-07-gpt-5",
        "benchmark": "aime-2025",
        "benchmark_name": "AIME 2025",
        "setting": "with-tools",
        "setting_name": "With tools",
        "score": 99.6,
        "unit": "%",
        "printed": "99.6%",
        "reported_by": "openai",
        "run_by": null,
        "label": "GPT-5 (python), With thinking",
        "source": "https://openai.com/index/introducing-gpt-5",
        "source_title": "Introducing GPT-5",
        "source_date": "2025-08-07",
        "where": "'Evaluations' section, bar chart 'AIME 2025 / Competition math', 3rd bar 'GPT-5 (python)', top (dark) segment",
        "own": true,
        "primary": true,
        "parallel": false,
        "notes": "with thinking; Python; pass@1; OpenAI says not to compare with-tools scores to models without tools",
        "caveat": "",
        "human_verified": true,
        "human_verified_on": "2026-10-18",
        "check": "human",
        "corroborated": true,
        "corroborated_on": "2026-10-19",
        "corroborated_by": [
          {
            "url": "https://www.anthropic.com/news/claude-sonnet-4-5",
            "name": "Claude Agent SDK",
            "title": "Introducing Claude Sonnet 4.5",
            "lab": "anthropic",
            "date": "2025-09-29",
            "printed": "99.6%",
            "where": "Benchmark table image, row 'High school math competition AIME 2025', column 'GPT-5', upper value '(python)'"
          },
          {
            "url": "https://www.kimi.ai/blog/kimi-k2-thinking",
            "name": "Kimi K2 Thinking",
            "title": "Introducing Kimi K2 Thinking",
            "lab": "moonshot",
            "date": "2025-11-06",
            "printed": "99.6",
            "where": "Full Evaluations table, row AIME 2025, w/ python (avg@16) ; column GPT-5 (High)"
          }
        ],
        "source_capture": "https://frontierlog.ai/captures/a16f4aa39159111d.png"
      },
      {
        "id": "gpt-5:aime-2025:with-tools:2",
        "model": "gpt-5",
        "model_name": "GPT-5",
        "model_date": "2025-08-07",
        "event": "https://frontierlog.ai/event/2025-08-07-gpt-5",
        "benchmark": "aime-2025",
        "benchmark_name": "AIME 2025",
        "setting": "with-tools",
        "setting_name": "With tools",
        "score": 71.0,
        "unit": "%",
        "printed": "71.0%",
        "reported_by": "openai",
        "run_by": null,
        "label": "GPT-5 (python), Without thinking",
        "source": "https://openai.com/index/introducing-gpt-5",
        "source_title": "Introducing GPT-5",
        "source_date": "2025-08-07",
        "where": "'Evaluations' section, bar chart 'AIME 2025 / Competition math', 3rd bar 'GPT-5 (python)', lower (light) segment",
        "own": true,
        "primary": false,
        "parallel": false,
        "notes": "without thinking; Python; pass@1",
        "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/a16f4aa39159111d.png"
      },
      {
        "id": "gpt-5:gpqa-diamond:default:1",
        "model": "gpt-5",
        "model_name": "GPT-5",
        "model_date": "2025-08-07",
        "event": "https://frontierlog.ai/event/2025-08-07-gpt-5",
        "benchmark": "gpqa-diamond",
        "benchmark_name": "GPQA Diamond",
        "setting": "default",
        "setting_name": "",
        "score": 85.7,
        "unit": "%",
        "printed": "85.7%",
        "reported_by": "openai",
        "run_by": null,
        "label": "GPT-5 (no tools), With thinking",
        "source": "https://openai.com/index/introducing-gpt-5",
        "source_title": "Introducing GPT-5",
        "source_date": "2025-08-07",
        "where": "'Evaluations' section, bar chart 'GPQA Diamond / PhD-level science questions', 4th bar 'GPT-5 (no tools)', top (dark, with thinking) value (page screenshot (job 20879 capture), part 05)",
        "own": true,
        "primary": true,
        "parallel": false,
        "notes": "with thinking; no tools; pass@1",
        "caveat": "",
        "human_verified": true,
        "human_verified_on": "2026-10-20",
        "check": "human",
        "corroborated": true,
        "corroborated_on": "2026-10-19",
        "corroborated_by": [
          {
            "url": "https://www.anthropic.com/news/claude-sonnet-4-5",
            "name": "Claude Agent SDK",
            "title": "Introducing Claude Sonnet 4.5",
            "lab": "anthropic",
            "date": "2025-09-29",
            "printed": "85.7%",
            "where": "Benchmark table image, row \"Graduate-level reasoning GPQA Diamond\", column \"GPT-5\"; thinking/tools not stated; footnote: OpenAI scores from GPT-5 post, GPT-5 for developers post, GPT-5 system card"
          },
          {
            "url": "https://openai.com/index/introducing-gpt-5-for-developers",
            "name": "OpenAI post",
            "title": "Introducing GPT-5 for developers",
            "lab": "openai",
            "date": "2025-08-07",
            "printed": "85.7%",
            "where": "Intelligence table, row 'GPQA diamond (no tools)', column 'GPT-5 (high)'"
          },
          {
            "url": "https://openai.com/index/gpt-5-1-for-developers",
            "name": "GPT-5.1 in the API",
            "title": "Introducing GPT-5.1 for developers",
            "lab": "openai",
            "date": "2025-11-13",
            "printed": "85.7%",
            "where": "Evaluation table, row 'GPQA Diamond (no tools)', column 'GPT-5 (high)'"
          }
        ],
        "source_capture": "https://frontierlog.ai/captures/b147a05c466ed2c8.png"
      },
      {
        "id": "gpt-5:gpqa-diamond:default:2",
        "model": "gpt-5",
        "model_name": "GPT-5",
        "model_date": "2025-08-07",
        "event": "https://frontierlog.ai/event/2025-08-07-gpt-5",
        "benchmark": "gpqa-diamond",
        "benchmark_name": "GPQA Diamond",
        "setting": "default",
        "setting_name": "",
        "score": 77.8,
        "unit": "%",
        "printed": "77.8%",
        "reported_by": "openai",
        "run_by": null,
        "label": "GPT-5 (no tools), Without thinking",
        "source": "https://openai.com/index/introducing-gpt-5",
        "source_title": "Introducing GPT-5",
        "source_date": "2025-08-07",
        "where": "'Evaluations' section, bar chart 'GPQA Diamond / PhD-level science questions', 4th bar 'GPT-5 (no tools)', lower (light, without thinking) segment (page screenshot (job 20879 capture), part 05)",
        "own": true,
        "primary": false,
        "parallel": false,
        "notes": "without thinking; no tools; pass@1",
        "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/b147a05c466ed2c8.png"
      },
      {
        "id": "gpt-5:gpqa-diamond:default:3",
        "model": "gpt-5",
        "model_name": "GPT-5",
        "model_date": "2025-08-07",
        "event": "https://frontierlog.ai/event/2025-08-07-gpt-5",
        "benchmark": "gpqa-diamond",
        "benchmark_name": "GPQA Diamond",
        "setting": "default",
        "setting_name": "",
        "score": 87.3,
        "unit": "%",
        "printed": "87.3%",
        "reported_by": "openai",
        "run_by": null,
        "label": "GPT-5 (python), With thinking",
        "source": "https://openai.com/index/introducing-gpt-5",
        "source_title": "Introducing GPT-5",
        "source_date": "2025-08-07",
        "where": "'Evaluations' section, bar chart 'GPQA Diamond / PhD-level science questions', 3rd bar 'GPT-5 (python)' (page screenshot (job 20879 capture), part 05)",
        "own": true,
        "primary": false,
        "parallel": false,
        "notes": "with thinking; Python; pass@1",
        "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/b147a05c466ed2c8.png"
      },
      {
        "id": "gpt-5:hle:no-tools:1",
        "model": "gpt-5",
        "model_name": "GPT-5",
        "model_date": "2025-08-07",
        "event": "https://frontierlog.ai/event/2025-08-07-gpt-5",
        "benchmark": "hle",
        "benchmark_name": "Humanity's Last Exam",
        "setting": "no-tools",
        "setting_name": "No tools",
        "score": 24.8,
        "unit": "%",
        "printed": "24.8%",
        "reported_by": "openai",
        "run_by": null,
        "label": "GPT-5 (no tools), With thinking",
        "source": "https://openai.com/index/introducing-gpt-5",
        "source_title": "Introducing GPT-5",
        "source_date": "2025-08-07",
        "where": "'Evaluations' section, bar chart 'Humanity's Last Exam (Full Set)*', 4th bar 'GPT-5 (no tools)', top (with thinking) value (page screenshot (job 20879 capture), part 05)",
        "own": true,
        "primary": true,
        "parallel": false,
        "notes": "with thinking; full set; pass@1",
        "caveat": "",
        "human_verified": true,
        "human_verified_on": "2026-10-20",
        "check": "human",
        "corroborated": true,
        "corroborated_on": "2026-10-19",
        "corroborated_by": [
          {
            "url": "https://www.kimi.ai/blog/kimi-k2-thinking",
            "name": "Kimi K2 Thinking",
            "title": "Introducing Kimi K2 Thinking",
            "lab": "moonshot",
            "date": "2025-11-06",
            "printed": "24.8",
            "where": "Footnote 3.b (prose, not the table): 'the official GPT-5 score on the HLE full set (no tools) is 24.8'. The table cell itself prints 26.3, which the footnote says is Scale.ai's text-only score."
          },
          {
            "url": "https://openai.com/index/introducing-gpt-5-for-developers",
            "name": "OpenAI post",
            "title": "Introducing GPT-5 for developers",
            "lab": "openai",
            "date": "2025-08-07",
            "printed": "24.8%",
            "where": "Intelligence table, row 'HLE (no tools)' (footnote 1: newer HLE version than the previous post), column 'GPT-5 (high)'"
          }
        ],
        "source_capture": "https://frontierlog.ai/captures/04f515e270a84d5f.png"
      },
      {
        "id": "gpt-5:hle:no-tools:2",
        "model": "gpt-5",
        "model_name": "GPT-5",
        "model_date": "2025-08-07",
        "event": "https://frontierlog.ai/event/2025-08-07-gpt-5",
        "benchmark": "hle",
        "benchmark_name": "Humanity's Last Exam",
        "setting": "no-tools",
        "setting_name": "No tools",
        "score": 6.3,
        "unit": "%",
        "printed": "6.3%",
        "reported_by": "openai",
        "run_by": null,
        "label": "GPT-5 (no tools), Without thinking",
        "source": "https://openai.com/index/introducing-gpt-5",
        "source_title": "Introducing GPT-5",
        "source_date": "2025-08-07",
        "where": "'Evaluations' section, bar chart 'Humanity's Last Exam (Full Set)*', 4th bar 'GPT-5 (no tools)', lower (light, without thinking) segment (page screenshot (job 20879 capture), part 05)",
        "own": true,
        "primary": false,
        "parallel": false,
        "notes": "without thinking; full set; pass@1",
        "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/04f515e270a84d5f.png"
      },
      {
        "id": "gpt-5:hle:with-tools:1",
        "model": "gpt-5",
        "model_name": "GPT-5",
        "model_date": "2025-08-07",
        "event": "https://frontierlog.ai/event/2025-08-07-gpt-5",
        "benchmark": "hle",
        "benchmark_name": "Humanity's Last Exam",
        "setting": "with-tools",
        "setting_name": "With tools",
        "score": 35.2,
        "unit": "%",
        "printed": "35.2%",
        "reported_by": "openai",
        "run_by": null,
        "label": "GPT-5 (python + search with blocklist)",
        "source": "https://openai.com/index/introducing-gpt-5",
        "source_title": "Introducing GPT-5",
        "source_date": "2025-08-07",
        "where": "'Evaluations' section, bar chart 'Humanity's Last Exam (Full Set)*', 3rd bar 'GPT-5 (python + search with blocklist)' (page screenshot (job 20879 capture), part 05)",
        "own": true,
        "primary": true,
        "parallel": false,
        "notes": "with thinking; Python and search with a blocklist; full set; pass@1",
        "caveat": "",
        "human_verified": true,
        "human_verified_on": "2026-10-20",
        "check": "human",
        "corroborated": true,
        "corroborated_on": "2026-10-19",
        "corroborated_by": [
          {
            "url": "https://www.kimi.ai/blog/kimi-k2-thinking",
            "name": "Kimi K2 Thinking",
            "title": "Introducing Kimi K2 Thinking",
            "lab": "moonshot",
            "date": "2025-11-06",
            "printed": "35.2",
            "where": "Footnote 3.b (prose): 'The GPT-5 ... on the HLE full set with tools are 35.2 ... from their official posts' (table cell shows 41.7, Moonshot's own text-only run)."
          },
          {
            "url": "https://huggingface.co/deepseek-ai/DeepSeek-V3.2/resolve/main/assets/paper.pdf",
            "name": "DeepSeek-V3.2 card",
            "title": "DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models",
            "lab": "deepseek",
            "date": "2025-12-01",
            "printed": "35.2",
            "where": "Table 2, 'Search Agent' section, row 'HLE (Pass@1)', column 'GPT-5 High'. No provenance footnote."
          },
          {
            "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": "35.2",
            "where": "Benchmark table, row 'HLE (w/ Tools)', column 'GPT-5 High'. No provenance footnote."
          }
        ],
        "source_capture": "https://frontierlog.ai/captures/04f515e270a84d5f.png"
      },
      {
        "id": "gpt-5:browsecomp:default:1",
        "model": "gpt-5",
        "model_name": "GPT-5",
        "model_date": "2025-08-07",
        "event": "https://frontierlog.ai/event/2025-08-07-gpt-5",
        "benchmark": "browsecomp",
        "benchmark_name": "BrowseComp",
        "setting": "default",
        "setting_name": "",
        "score": 54.9,
        "unit": "%",
        "printed": "54.9%",
        "reported_by": "openai",
        "run_by": null,
        "label": "GPT-5, With thinking",
        "source": "https://openai.com/index/introducing-gpt-5",
        "source_title": "Introducing GPT-5",
        "source_date": "2025-08-07",
        "where": "'Instruction following and agentic tool use' section, bar chart 'BrowseComp / Agentic search & browsing', 1st bar 'GPT-5' (page screenshot (job 20879 capture), part 06)",
        "own": true,
        "primary": true,
        "parallel": false,
        "notes": "with thinking; pass@1",
        "caveat": "",
        "human_verified": true,
        "human_verified_on": "2026-10-20",
        "check": "human",
        "corroborated": true,
        "corroborated_on": "2026-10-19",
        "corroborated_by": [
          {
            "url": "https://www.kimi.ai/blog/kimi-k2-thinking",
            "name": "Kimi K2 Thinking",
            "title": "Introducing Kimi K2 Thinking",
            "lab": "moonshot",
            "date": "2025-11-06",
            "printed": "54.9",
            "where": "Agentic search table, row 'BrowseComp' (w/ tools), column 'GPT-5 (High)'. Unstarred = quoted from OpenAI per footnote 3.a."
          },
          {
            "url": "https://huggingface.co/deepseek-ai/DeepSeek-V3.2/resolve/main/assets/paper.pdf",
            "name": "DeepSeek-V3.2 card",
            "title": "DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models",
            "lab": "deepseek",
            "date": "2025-12-01",
            "printed": "54.9",
            "where": "Table 2, row 'BrowseComp (Pass@1)', column 'GPT-5 High'. No provenance footnote."
          },
          {
            "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": "54.9",
            "where": "Benchmark table, row 'BrowseComp', column 'GPT-5 High'. No provenance footnote."
          }
        ],
        "source_capture": "https://frontierlog.ai/captures/c35813e009a6d32a.png"
      },
      {
        "id": "gpt-5:swe-bench-verified:default:2",
        "model": "gpt-5",
        "model_name": "GPT-5",
        "model_date": "2025-08-07",
        "event": "https://frontierlog.ai/event/2025-08-07-gpt-5",
        "benchmark": "swe-bench-verified",
        "benchmark_name": "SWE-bench Verified",
        "setting": "default",
        "setting_name": "",
        "score": 72.8,
        "unit": "%",
        "printed": "72.8%",
        "reported_by": "openai",
        "run_by": null,
        "label": "GPT-5 (high)",
        "source": "https://openai.com/index/gpt-5-1-for-developers",
        "source_title": "Introducing GPT-5.1 for developers",
        "source_date": "2025-11-13",
        "where": "evaluation table, row 'SWE-bench Verified (all 500 problems)', column 'GPT-5 (high)'",
        "own": false,
        "primary": false,
        "parallel": false,
        "notes": "high reasoning effort; all 500 problems",
        "caveat": "",
        "human_verified": false,
        "human_verified_on": null,
        "check": "corroborated",
        "corroborated": true,
        "corroborated_on": "2026-10-20",
        "corroborated_by": [
          {
            "url": "https://www.anthropic.com/news/claude-sonnet-4-5",
            "name": "Claude Agent SDK",
            "title": "Introducing Claude Sonnet 4.5",
            "lab": "anthropic",
            "date": "2025-09-29",
            "printed": "72.8%",
            "where": "main benchmark table image, row 'Agentic coding / SWE-bench Verified', column 'GPT-5', value labelled 'GPT-5'"
          },
          {
            "url": "https://www.anthropic.com/news/claude-haiku-4-5",
            "name": "Claude Haiku 4.5",
            "title": "Introducing Claude Haiku 4.5",
            "lab": "anthropic",
            "date": "2025-10-15",
            "printed": "72.8%",
            "where": "main benchmark table image, row 'Agentic coding / SWE-bench Verified', column 'GPT-5', value labelled 'GPT-5 (high)'"
          }
        ],
        "source_capture": "https://frontierlog.ai/captures/5fc625949d807803.png"
      }
    ]
  }
}