{
  "api_version": "1",
  "updated": "2026-10-20",
  "latest_event": "2026-10-18",
  "rights": "© Frontier Log, all rights reserved.",
  "benchmark": {
    "slug": "tau2-bench",
    "name": "τ²-bench",
    "family": "tau-bench",
    "unit": "%",
    "higher_is_better": true,
    "status": "retired",
    "active": false,
    "superseded_by": null,
    "benchmark_url": "https://github.com/sierra-research/tau2-bench",
    "introduced": "2025-06-09",
    "creator": "Sierra",
    "introduced_source": "https://arxiv.org/abs/2506.07982",
    "settings": [
      {
        "setting": "telecom",
        "name": "Telecom",
        "url": "https://frontierlog.ai/benchmark/tau2-bench",
        "results": 4
      },
      {
        "setting": "retail",
        "name": "Retail",
        "url": "https://frontierlog.ai/benchmark/tau2-bench/retail",
        "results": 3
      },
      {
        "setting": "airline",
        "name": "Airline",
        "url": "https://frontierlog.ai/benchmark/tau2-bench/airline",
        "results": 3
      }
    ]
  },
  "total": 3,
  "results": [
    {
      "id": "nemotron-3-nano-30b-a3b:tau2-bench:airline:1",
      "model": "nemotron-3-nano-30b-a3b",
      "model_name": "Nemotron 3 Nano 30B-A3B",
      "model_date": "2025-12-15",
      "event": "https://frontierlog.ai/event/2025-12-15-nemotron-3",
      "benchmark": "tau2-bench",
      "benchmark_name": "τ²-bench",
      "setting": "airline",
      "setting_name": "Airline",
      "score": 48.0,
      "unit": "%",
      "printed": "48.0",
      "reported_by": "nvidia",
      "label": "NVIDIA-Nemotron-3-Nano-30B-A3B-BF16",
      "source": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-FP8/blob/fa5699ff0bb9da723217af52488c0f9653512ffb/README.md",
      "source_title": "nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-FP8 model card (launch-day revision)",
      "source_date": "2025-12-15",
      "where": "'Reasoning Benchmark Evaluations' table, row 'TauBench V2 (Airline)', column 'NVIDIA-Nemotron-3-Nano-30B-A3B-BF16'",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "",
      "caveat": "",
      "human_verified": false,
      "human_verified_on": null,
      "check": "corroborated",
      "corroborated": true,
      "corroborated_on": "2026-10-19",
      "corroborated_by": [
        {
          "url": "https://research.nvidia.com/labs/nemotron/files/NVIDIA-Nemotron-3-Nano-Technical-Report.pdf",
          "name": "Nemotron 3 card",
          "title": "Nemotron 3 Nano: Open, Efficient Mixture-of-Experts Hybrid Mamba-Transformer Model for Agentic Reasoning",
          "lab": "nvidia",
          "date": "2025-12-15",
          "printed": "48.00",
          "where": "Table 3, TauBench V2 Airline row, N-3-Nano column"
        },
        {
          "url": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16/blob/bf77c3174f68ad409e1c2aa60daeb46e32d1c606/README.md",
          "name": "Nemotron 3 card",
          "title": "nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16 model card (revision of 2026-08-24)",
          "lab": "nvidia",
          "date": "2025-12-15",
          "printed": "48.0",
          "where": "Benchmarks table, 'Nemotron-3-Nano' (first, BF16) column, TauBench V2 (Airline) row"
        },
        {
          "url": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-NVFP4/blob/6efb4a2a1c1fa277ce7b3df7a1416255011b1c99/README.md",
          "name": "Nemotron 3 card",
          "title": "nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-NVFP4 model card (revision of 2026-08-24)",
          "lab": "nvidia",
          "date": "2025-12-15",
          "printed": "48.0",
          "where": "accuracy table (BF16 | FP8 | NVFP4), first (BF16) column, TauBench V2 (Airline) row"
        }
      ]
    },
    {
      "id": "nemotron-3-super-120b-a12b:tau2-bench:airline:1",
      "model": "nemotron-3-super-120b-a12b",
      "model_name": "Nemotron 3 Super 120B-A12B",
      "model_date": "2026-03-11",
      "event": "https://frontierlog.ai/event/2026-03-11-nemotron-3-super",
      "benchmark": "tau2-bench",
      "benchmark_name": "τ²-bench",
      "setting": "airline",
      "setting_name": "Airline",
      "score": 56.25,
      "unit": "%",
      "printed": "56.25",
      "reported_by": "nvidia",
      "label": "Nemotron-3-Super",
      "source": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-FP8/blob/1408abd63ba6ba1c37a78b94d332dac35da45370/README.md",
      "source_title": "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-FP8 model card (launch-day revision)",
      "source_date": "2026-03-11",
      "where": "benchmark table under the accuracy chart, row 'TauBench V2 / Airline', column 'Nemotron-3-Super'",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "default prompt",
      "caveat": "",
      "human_verified": false,
      "human_verified_on": null,
      "check": "corroborated",
      "corroborated": true,
      "corroborated_on": "2026-10-19",
      "corroborated_by": [
        {
          "url": "https://research.nvidia.com/labs/nemotron/files/NVIDIA-Nemotron-3-Super-Technical-Report.pdf",
          "name": "Nemotron 3 Super card",
          "title": "Nemotron 3 Super: Open, Efficient Mixture-of-Experts Hybrid Mamba-Transformer Model for Agentic Reasoning",
          "lab": "nvidia",
          "date": "2026-03-11",
          "printed": "56.25",
          "where": "Table 5, TauBench V2 Airline row, N-3-Super column"
        },
        {
          "url": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-BF16/blob/2dc98e2afe4face0e4ce40972a915c45368bd34a/README.md",
          "name": "Nemotron 3 Super card",
          "title": "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-BF16 model card (revision of 2026-08-25)",
          "lab": "nvidia",
          "date": "2026-03-11",
          "printed": "56.25",
          "where": "Benchmarks table, 'Nemotron 3 Super' column, TauBench V2 Airline row"
        },
        {
          "url": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-NVFP4/blob/ff433f5493e25d631c9f12b5d55c674229923d02/README.md",
          "name": "Nemotron 3 Super card",
          "title": "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-NVFP4 model card (revision of 2026-08-24)",
          "lab": "nvidia",
          "date": "2026-03-11",
          "printed": "56.25",
          "where": "accuracy table, 'Nemotron-3-Super' (BF16) column, TauBench V2 Airline row"
        }
      ]
    },
    {
      "id": "nemotron-3-ultra-550b-a55b:tau2-bench:airline:1",
      "model": "nemotron-3-ultra-550b-a55b",
      "model_name": "Nemotron 3 Ultra 550B-A55B",
      "model_date": "2026-06-04",
      "event": "https://frontierlog.ai/event/2026-06-04-nemotron-3-ultra",
      "benchmark": "tau2-bench",
      "benchmark_name": "τ²-bench",
      "setting": "airline",
      "setting_name": "Airline",
      "score": 79.75,
      "unit": "%",
      "printed": "79.75",
      "reported_by": "nvidia",
      "label": "N3-Ultra 550B-A55B",
      "source": "https://research.nvidia.com/labs/nemotron/files/NVIDIA-Nemotron-3.5-Super-and-Lightning-Technical-Report.pdf",
      "source_title": "Nemotron 3.5 Super and Lightning: Open, Efficient Mixture-of-Experts Hybrid Mamba-Transformer Models for Agentic AI",
      "source_date": "2026-10-15",
      "where": "Table 10 (PDF page 23), row 'Tau2 Airline', column 'N3-Ultra 550B-A55B' (second column)",
      "own": false,
      "primary": true,
      "parallel": false,
      "notes": "average of 8 trials; Qwen3-235B-A22B-Instruct-2507 user simulator",
      "caveat": "",
      "human_verified": false,
      "human_verified_on": null,
      "check": "corroborated",
      "corroborated": true,
      "corroborated_on": "2026-10-19",
      "corroborated_by": [
        {
          "url": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-BF16/blob/1f8953f7f0935358e3605150d22bc15df062061c/README.md",
          "name": "Nemotron 3.5 Super VL and Bolt card",
          "title": "nvidia/NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-BF16 model card (revision of 2026-10-16)",
          "lab": "nvidia",
          "date": "2026-10-15",
          "printed": "79.8",
          "where": "evaluation table, 'Nemotron 3 Ultra' column, tau2-bench airline row"
        }
      ]
    }
  ]
}