{
  "api_version": "1",
  "updated": "2026-10-26",
  "latest_event": "2026-10-24",
  "rights": "© Frontier Log, all rights reserved.",
  "model": {
    "slug": "glm-5-2",
    "name": "GLM-5.2",
    "lab": "zai",
    "date": "2026-06-16",
    "flagship": false,
    "event": "https://frontierlog.ai/event/2026-06-16-glm-5-2",
    "releases": [],
    "variant_of": null,
    "made_by": "Z.ai",
    "url": "https://frontierlog.ai/model/glm-5-2",
    "results": [
      {
        "id": "glm-5-2:hle:no-tools:1",
        "model": "glm-5-2",
        "model_name": "GLM-5.2",
        "model_date": "2026-06-16",
        "event": "https://frontierlog.ai/event/2026-06-16-glm-5-2",
        "benchmark": "hle",
        "benchmark_name": "Humanity's Last Exam",
        "setting": "no-tools",
        "setting_name": "No tools",
        "score": 40.5,
        "unit": "%",
        "printed": "40.5",
        "reported_by": "zai",
        "run_by": null,
        "label": "GLM-5.2",
        "source": "https://z.ai/blog/glm-5.2",
        "source_title": "GLM-5.2: Built for Long-Horizon Tasks",
        "source_date": "2026-06-16",
        "where": "'Full Benchmark Table', row 'HLE', column 'GLM-5.2'",
        "own": true,
        "primary": true,
        "parallel": false,
        "notes": "text-only subset; GPT-5.5 judge; up to 163,840 output tokens",
        "caveat": "",
        "human_verified": true,
        "human_verified_on": "2026-10-19",
        "check": "human",
        "corroborated": false,
        "corroborated_on": null,
        "corroborated_by": [],
        "source_capture": "https://frontierlog.ai/captures/112d3aba81edfc46.png"
      },
      {
        "id": "glm-5-2:hle:with-tools:1",
        "model": "glm-5-2",
        "model_name": "GLM-5.2",
        "model_date": "2026-06-16",
        "event": "https://frontierlog.ai/event/2026-06-16-glm-5-2",
        "benchmark": "hle",
        "benchmark_name": "Humanity's Last Exam",
        "setting": "with-tools",
        "setting_name": "With tools",
        "score": 54.7,
        "unit": "%",
        "printed": "54.7",
        "reported_by": "zai",
        "run_by": null,
        "label": "GLM-5.2",
        "source": "https://z.ai/blog/glm-5.2",
        "source_title": "GLM-5.2: Built for Long-Horizon Tasks",
        "source_date": "2026-06-16",
        "where": "'Full Benchmark Table', row 'HLE w/ Tools', column 'GLM-5.2'",
        "own": true,
        "primary": true,
        "parallel": false,
        "notes": "text-only subset; 300,000-token context, no context management",
        "caveat": "",
        "human_verified": true,
        "human_verified_on": "2026-10-19",
        "check": "human",
        "corroborated": true,
        "corroborated_on": "2026-10-19",
        "corroborated_by": [
          {
            "url": "https://thinkingmachines.ai/news/introducing-inkling/",
            "name": "Inkling",
            "title": "Inkling: Our Open-Weights Model",
            "lab": "thinking-machines",
            "date": "2026-07-15",
            "printed": "54.7%",
            "where": "'Benchmarking Inkling' main table, row 'HLE with tools', column 'GLM 5.2'"
          },
          {
            "url": "https://z.ai/blog/glm-5.3",
            "name": "GLM-5.3",
            "title": "GLM-5.3: Frontier Coding with Emergent Cyber Capabilities",
            "lab": "zai",
            "date": "2026-08-14",
            "printed": "54.7",
            "where": "'Performance across comparison models' table, row 'HLE w/ Tools', column 'GLM-5.2'"
          }
        ],
        "source_capture": "https://frontierlog.ai/captures/112d3aba81edfc46.png"
      },
      {
        "id": "glm-5-2:gpqa-diamond:default:1",
        "model": "glm-5-2",
        "model_name": "GLM-5.2",
        "model_date": "2026-06-16",
        "event": "https://frontierlog.ai/event/2026-06-16-glm-5-2",
        "benchmark": "gpqa-diamond",
        "benchmark_name": "GPQA Diamond",
        "setting": "default",
        "setting_name": "",
        "score": 91.2,
        "unit": "%",
        "printed": "91.2",
        "reported_by": "zai",
        "run_by": null,
        "label": "GLM-5.2",
        "source": "https://z.ai/blog/glm-5.2",
        "source_title": "GLM-5.2: Built for Long-Horizon Tasks",
        "source_date": "2026-06-16",
        "where": "'Full Benchmark Table', row 'GPQA-Diamond', column 'GLM-5.2'",
        "own": true,
        "primary": true,
        "parallel": false,
        "notes": "up to 163,840 output tokens",
        "caveat": "",
        "human_verified": true,
        "human_verified_on": "2026-10-19",
        "check": "human",
        "corroborated": true,
        "corroborated_on": "2026-10-19",
        "corroborated_by": [
          {
            "url": "https://www.kimi.ai/blog/kimi-k3",
            "name": "Kimi K3",
            "title": "Kimi K3: Open Frontier Intelligence",
            "lab": "moonshot",
            "date": "2026-07-16",
            "printed": "91.2",
            "where": "Full Benchmark Table, row GPQA-Diamond ; column GLM-5.2 (max)"
          }
        ],
        "source_capture": "https://frontierlog.ai/captures/112d3aba81edfc46.png"
      },
      {
        "id": "glm-5-2:swe-bench-pro:default:1",
        "model": "glm-5-2",
        "model_name": "GLM-5.2",
        "model_date": "2026-06-16",
        "event": "https://frontierlog.ai/event/2026-06-16-glm-5-2",
        "benchmark": "swe-bench-pro",
        "benchmark_name": "SWE-bench Pro",
        "setting": "default",
        "setting_name": "",
        "score": 62.1,
        "unit": "%",
        "printed": "62.1",
        "reported_by": "zai",
        "run_by": null,
        "label": "GLM-5.2",
        "source": "https://z.ai/blog/glm-5.2",
        "source_title": "GLM-5.2: Built for Long-Horizon Tasks",
        "source_date": "2026-06-16",
        "where": "'Full Benchmark Table', row 'SWE-bench Pro', column 'GLM-5.2'",
        "own": true,
        "primary": true,
        "parallel": false,
        "notes": "OpenHands with a tailored prompt; 400K context",
        "caveat": "The source doesn't say whether this is the public or the commercial set",
        "human_verified": true,
        "human_verified_on": "2026-10-19",
        "check": "human",
        "corroborated": true,
        "corroborated_on": "2026-10-19",
        "corroborated_by": [
          {
            "url": "https://thinkingmachines.ai/news/introducing-inkling/",
            "name": "Inkling",
            "title": "Inkling: Our Open-Weights Model",
            "lab": "thinking-machines",
            "date": "2026-07-15",
            "printed": "62.1%",
            "where": "'Benchmarking Inkling' main table, row 'SWEBench Pro Public', column 'GLM 5.2'"
          }
        ],
        "source_capture": "https://frontierlog.ai/captures/112d3aba81edfc46.png"
      },
      {
        "id": "glm-5-2:deepswe-1-1:default:1",
        "model": "glm-5-2",
        "model_name": "GLM-5.2",
        "model_date": "2026-06-16",
        "event": "https://frontierlog.ai/event/2026-06-16-glm-5-2",
        "benchmark": "deepswe-1-1",
        "benchmark_name": "DeepSWE v1.1",
        "setting": "default",
        "setting_name": "",
        "score": 46.2,
        "unit": "%",
        "printed": "46.2",
        "reported_by": "zai",
        "run_by": null,
        "label": "GLM-5.2",
        "source": "https://z.ai/blog/glm-5.2",
        "source_title": "GLM-5.2: Built for Long-Horizon Tasks",
        "source_date": "2026-06-16",
        "where": "'Full Benchmark Table', row 'DeepSWE', column 'GLM-5.2'",
        "own": true,
        "primary": true,
        "parallel": false,
        "notes": "mini-swe-agent harness; 2h timeout; 400K context",
        "caveat": "",
        "human_verified": true,
        "human_verified_on": "2026-10-19",
        "check": "human",
        "corroborated": true,
        "corroborated_on": "2026-10-19",
        "corroborated_by": [
          {
            "url": "https://www.kimi.ai/blog/kimi-k3",
            "name": "Kimi K3",
            "title": "Kimi K3: Open Frontier Intelligence",
            "lab": "moonshot",
            "date": "2026-07-16",
            "printed": "46.2",
            "where": "Full Benchmark Table, row DeepSWE (footnote: DeepSWE v1.1 tasks) ; column GLM-5.2 (max)"
          },
          {
            "url": "https://z.ai/blog/glm-5.3",
            "name": "GLM-5.3",
            "title": "GLM-5.3: Frontier Coding with Emergent Cyber Capabilities",
            "lab": "zai",
            "date": "2026-08-14",
            "printed": "46.2",
            "where": "Prose: 'from 46.2 to 66.9 on DeepSWE v1.1' (GLM-5.2)"
          }
        ],
        "source_capture": "https://frontierlog.ai/captures/112d3aba81edfc46.png"
      },
      {
        "id": "glm-5-2:terminal-bench-2-1:default:1",
        "model": "glm-5-2",
        "model_name": "GLM-5.2",
        "model_date": "2026-06-16",
        "event": "https://frontierlog.ai/event/2026-06-16-glm-5-2",
        "benchmark": "terminal-bench-2-1",
        "benchmark_name": "Terminal-Bench 2.1",
        "setting": "default",
        "setting_name": "",
        "score": 81.0,
        "unit": "%",
        "printed": "81.0",
        "reported_by": "zai",
        "run_by": null,
        "label": "GLM-5.2",
        "source": "https://z.ai/blog/glm-5.2",
        "source_title": "GLM-5.2: Built for Long-Horizon Tasks",
        "source_date": "2026-06-16",
        "where": "'Full Benchmark Table', row 'Terminal Bench 2.1 Terminus-2', column 'GLM-5.2'",
        "own": true,
        "primary": true,
        "parallel": false,
        "notes": "Terminus 2 harness; 4h timeout; 256K context",
        "caveat": "",
        "human_verified": true,
        "human_verified_on": "2026-10-19",
        "check": "human",
        "corroborated": true,
        "corroborated_on": "2026-10-19",
        "corroborated_by": [
          {
            "url": "https://z.ai/blog/glm-5.3",
            "name": "GLM-5.3",
            "title": "GLM-5.3: Frontier Coding with Emergent Cyber Capabilities",
            "lab": "zai",
            "date": "2026-08-14",
            "printed": "81.0",
            "where": "'Performance across comparison models' table, row 'Terminal Bench 2.1 (Claude Code harness per footnote)', column 'GLM-5.2'"
          }
        ],
        "source_capture": "https://frontierlog.ai/captures/112d3aba81edfc46.png"
      },
      {
        "id": "glm-5-2:terminal-bench-2-1:default:2",
        "model": "glm-5-2",
        "model_name": "GLM-5.2",
        "model_date": "2026-06-16",
        "event": "https://frontierlog.ai/event/2026-06-16-glm-5-2",
        "benchmark": "terminal-bench-2-1",
        "benchmark_name": "Terminal-Bench 2.1",
        "setting": "default",
        "setting_name": "",
        "score": 82.7,
        "unit": "%",
        "printed": "82.7",
        "reported_by": "zai",
        "run_by": null,
        "label": "GLM-5.2",
        "source": "https://z.ai/blog/glm-5.2",
        "source_title": "GLM-5.2: Built for Long-Horizon Tasks",
        "source_date": "2026-06-16",
        "where": "'Full Benchmark Table', row 'Terminal Bench 2.1 Best Reported Harness', column 'GLM-5.2'",
        "own": true,
        "primary": false,
        "parallel": false,
        "notes": "Claude Code harness; no time limit; averaged over 5 runs",
        "caveat": "",
        "human_verified": true,
        "human_verified_on": "2026-10-19",
        "check": "human",
        "corroborated": true,
        "corroborated_on": "2026-10-19",
        "corroborated_by": [
          {
            "url": "https://www.kimi.ai/blog/kimi-k3",
            "name": "Kimi K3",
            "title": "Kimi K3: Open Frontier Intelligence",
            "lab": "moonshot",
            "date": "2026-07-16",
            "printed": "82.7",
            "where": "Full Benchmark Table, row Terminal Bench 2.1 ; column GLM-5.2 (max)"
          },
          {
            "url": "https://thinkingmachines.ai/news/introducing-inkling/",
            "name": "Inkling",
            "title": "Inkling: Our Open-Weights Model",
            "lab": "thinking-machines",
            "date": "2026-07-15",
            "printed": "82.7%",
            "where": "'Benchmarking Inkling' main table, row 'Terminal Bench 2.1* Best Harness', column 'GLM 5.2'"
          }
        ],
        "source_capture": "https://frontierlog.ai/captures/112d3aba81edfc46.png"
      },
      {
        "id": "glm-5-2:mcp-atlas:default:1",
        "model": "glm-5-2",
        "model_name": "GLM-5.2",
        "model_date": "2026-06-16",
        "event": "https://frontierlog.ai/event/2026-06-16-glm-5-2",
        "benchmark": "mcp-atlas",
        "benchmark_name": "MCP Atlas",
        "setting": "default",
        "setting_name": "",
        "score": 76.8,
        "unit": "%",
        "printed": "76.8",
        "reported_by": "zai",
        "run_by": null,
        "label": "GLM-5.2",
        "source": "https://z.ai/blog/glm-5.2",
        "source_title": "GLM-5.2: Built for Long-Horizon Tasks",
        "source_date": "2026-06-16",
        "where": "'Full Benchmark Table', row 'MCP-Atlas Public Set', column 'GLM-5.2'",
        "own": true,
        "primary": false,
        "parallel": false,
        "notes": "500-task public set; thinking mode; 10-minute timeout per task; Gemini 3 Pro judge",
        "caveat": "",
        "human_verified": true,
        "human_verified_on": "2026-10-19",
        "check": "human",
        "corroborated": false,
        "corroborated_on": null,
        "corroborated_by": [],
        "source_capture": "https://frontierlog.ai/captures/112d3aba81edfc46.png"
      },
      {
        "id": "glm-5-2:mcp-atlas:default:2",
        "model": "glm-5-2",
        "model_name": "GLM-5.2",
        "model_date": "2026-06-16",
        "event": "https://frontierlog.ai/event/2026-06-16-glm-5-2",
        "benchmark": "mcp-atlas",
        "benchmark_name": "MCP Atlas",
        "setting": "default",
        "setting_name": "",
        "score": 77.0,
        "unit": "%",
        "printed": "77.0",
        "reported_by": "zai",
        "run_by": null,
        "label": "GLM-5.2",
        "source": "https://z.ai/blog/glm-5.2",
        "source_title": "GLM-5.2: Built for Long-Horizon Tasks",
        "source_date": "2026-06-16",
        "where": "'LLM Performance Evaluation' chart image (eight panels), panel 'MCP-Atlas', first bar 'GLM-5.2', value label",
        "own": true,
        "primary": true,
        "parallel": false,
        "notes": "maximum thinking effort; the post's chart, where its table prints 76.8",
        "caveat": "",
        "human_verified": true,
        "human_verified_on": "2026-10-19",
        "check": "human",
        "corroborated": false,
        "corroborated_on": null,
        "corroborated_by": [],
        "source_image": "https://frontierlog.ai/sources/bfcfa600d8d45ed8.png",
        "source_image_original": "https://z-cdn-media.chatglm.cn/prompts-rich-media-resources/5.2-blog/20260617-012836.png"
      },
      {
        "id": "glm-5-2:automationbench:default:1",
        "model": "glm-5-2",
        "model_name": "GLM-5.2",
        "model_date": "2026-06-16",
        "event": "https://frontierlog.ai/event/2026-06-16-glm-5-2",
        "benchmark": "automationbench",
        "benchmark_name": "AutomationBench",
        "setting": "default",
        "setting_name": "",
        "score": 26.2,
        "unit": "%",
        "printed": "26.2",
        "reported_by": "zai",
        "run_by": null,
        "label": "GLM-5.2",
        "source": "https://z.ai/blog/glm-5.3",
        "source_title": "GLM-5.3: Frontier Coding with Emergent Cyber Capabilities",
        "source_date": "2026-08-14",
        "where": "'Performance across comparison models' table, row 'AutomationBench v1.0.6', column 'GLM-5.2'",
        "own": false,
        "primary": true,
        "parallel": false,
        "notes": "AutomationBench v1.0.6",
        "caveat": "",
        "human_verified": true,
        "human_verified_on": "2026-10-19",
        "check": "human",
        "corroborated": false,
        "corroborated_on": null,
        "corroborated_by": [],
        "source_capture": "https://frontierlog.ai/captures/86a1ebfddd94c33e.png"
      }
    ]
  }
}