{
  "schemaVersion": 1,
  "benchmark": {
    "id": "open-asr-ami-cleaned",
    "name": "Open ASR · meeting transcription",
    "version": "AMI-Cleaned · English test · 2026-09-04",
    "category": "audio",
    "question": "Which open-weight speech models make fewer errors in English meetings?",
    "summary": "Ten selected configurations on the same cleaned meeting-transcription test. These are publisher-reported results with different inference pipelines.",
    "source": {
      "name": "Hugging Face Open ASR Leaderboard",
      "url": "https://huggingface.co/datasets/hf-audio/open-asr-leaderboard-results/blob/ba5712d5ace8f785fa0daae1aecea8561ecd87c9/english_short_latest.csv",
      "methodologyUrl": "https://github.com/huggingface/open_asr_leaderboard"
    },
    "measure": "Word error rate (WER), %; lower is better. Counts substituted, missing, and extra words.",
    "comparisonRule": "Compare only the AMI-Cleaned English test in this September 4, 2026 snapshot. Keep the model version and publisher scoring protocol fixed.",
    "limitations": [
      "A selected open-weight comparison, not the full leaderboard or a top-ten list.",
      "Inference pipelines differ. The published rows do not identify exact run dates, decoding settings, checkpoint revisions, or execution records.",
      "Short speech clips do not test whole-meeting speaker attribution, punctuation quality, other languages, or live response time.",
      "No uncertainty is published; small differences do not establish a reliable winner. WER can exceed 100% when extra words are inserted."
    ],
    "coverage": "charted",
    "tags": [
      "audio",
      "speech",
      "transcription",
      "ASR",
      "WER",
      "English",
      "meetings",
      "AMI",
      "open weights",
      "Whisper",
      "Parakeet",
      "Cohere",
      "Qwen",
      "Granite",
      "Voxtral"
    ]
  },
  "dataset": {
    "benchmarkId": "open-asr-ami-cleaned",
    "version": "AMI-Cleaned · English test · 2026-09-04",
    "score": {
      "label": "Word error rate",
      "unit": "% WER",
      "direction": "lower",
      "minimum": 0
    },
    "source": {
      "name": "Hugging Face Open ASR Leaderboard",
      "url": "https://huggingface.co/datasets/hf-audio/open-asr-leaderboard-results/blob/ba5712d5ace8f785fa0daae1aecea8561ecd87c9/english_short_latest.csv",
      "revision": "ba5712d5ace8f785fa0daae1aecea8561ecd87c9",
      "retrievedAt": "2026-09-09T01:53:06.273Z"
    },
    "observedAt": "2026-09-04",
    "evidenceLabel": "Open ASR owner-reported · selected open-weight configurations",
    "configurationLabel": "Published model configuration",
    "comparabilityNote": "Same AMI-Cleaned English test and publisher scoring protocol. Inference pipelines differ; exact per-run configurations are not published. Lower word-error rate is better. This is not a speed, speaker-attribution, or overall audio-quality ranking.",
    "points": [
      {
        "id": "granite-speech-4-1-2b-nar",
        "label": "Granite Speech 4.1 2B NAR",
        "model": "Granite Speech 4.1 2B NAR",
        "provider": "IBM",
        "harness": "Open ASR published configuration",
        "effort": null,
        "score": 6.96,
        "costUsd": null,
        "uncertainty": null,
        "sourceUrl": "https://huggingface.co/datasets/hf-audio/open-asr-leaderboard-results/blob/ba5712d5ace8f785fa0daae1aecea8561ecd87c9/english_short_latest.csv",
        "details": [
          {
            "label": "Source model ID",
            "value": "ibm-granite/granite-speech-4.1-2b-nar"
          },
          {
            "label": "Test",
            "value": "AMI-Cleaned · English · test split"
          },
          {
            "label": "Scoring",
            "value": "Publisher English normalization and compound-aware word-error rate"
          },
          {
            "label": "Exact execution configuration",
            "value": "Not published with this result; current launcher documentation is not a run receipt"
          },
          {
            "label": "Source snapshot",
            "value": "September 4, 2026; individual run date not published"
          }
        ]
      },
      {
        "id": "cohere-transcribe-03-2026",
        "label": "Cohere Transcribe 03-2026",
        "model": "Cohere Transcribe 03-2026",
        "provider": "Cohere",
        "harness": "Open ASR published configuration",
        "effort": null,
        "score": 7.02,
        "costUsd": null,
        "uncertainty": null,
        "sourceUrl": "https://huggingface.co/datasets/hf-audio/open-asr-leaderboard-results/blob/ba5712d5ace8f785fa0daae1aecea8561ecd87c9/english_short_latest.csv",
        "details": [
          {
            "label": "Source model ID",
            "value": "CohereLabs/cohere-transcribe-03-2026"
          },
          {
            "label": "Test",
            "value": "AMI-Cleaned · English · test split"
          },
          {
            "label": "Scoring",
            "value": "Publisher English normalization and compound-aware word-error rate"
          },
          {
            "label": "Exact execution configuration",
            "value": "Not published with this result; current launcher documentation is not a run receipt"
          },
          {
            "label": "Source snapshot",
            "value": "September 4, 2026; individual run date not published"
          }
        ]
      },
      {
        "id": "canary-qwen-2-5b",
        "label": "Canary-Qwen 2.5B",
        "model": "Canary-Qwen 2.5B",
        "provider": "NVIDIA",
        "harness": "Open ASR published configuration",
        "effort": null,
        "score": 7.91,
        "costUsd": null,
        "uncertainty": null,
        "sourceUrl": "https://huggingface.co/datasets/hf-audio/open-asr-leaderboard-results/blob/ba5712d5ace8f785fa0daae1aecea8561ecd87c9/english_short_latest.csv",
        "details": [
          {
            "label": "Source model ID",
            "value": "nvidia/canary-qwen-2.5b"
          },
          {
            "label": "Test",
            "value": "AMI-Cleaned · English · test split"
          },
          {
            "label": "Scoring",
            "value": "Publisher English normalization and compound-aware word-error rate"
          },
          {
            "label": "Exact execution configuration",
            "value": "Not published with this result; current launcher documentation is not a run receipt"
          },
          {
            "label": "Source snapshot",
            "value": "September 4, 2026; individual run date not published"
          }
        ]
      },
      {
        "id": "qwen3-asr-1-7b-hf",
        "label": "Qwen3-ASR 1.7B HF",
        "model": "Qwen3-ASR 1.7B HF",
        "provider": "Alibaba",
        "harness": "Open ASR published configuration",
        "effort": null,
        "score": 8.31,
        "costUsd": null,
        "uncertainty": null,
        "sourceUrl": "https://huggingface.co/datasets/hf-audio/open-asr-leaderboard-results/blob/ba5712d5ace8f785fa0daae1aecea8561ecd87c9/english_short_latest.csv",
        "details": [
          {
            "label": "Source model ID",
            "value": "Qwen/Qwen3-ASR-1.7B-hf"
          },
          {
            "label": "Test",
            "value": "AMI-Cleaned · English · test split"
          },
          {
            "label": "Scoring",
            "value": "Publisher English normalization and compound-aware word-error rate"
          },
          {
            "label": "Exact execution configuration",
            "value": "Not published with this result; current launcher documentation is not a run receipt"
          },
          {
            "label": "Source snapshot",
            "value": "September 4, 2026; individual run date not published"
          }
        ]
      },
      {
        "id": "parakeet-tdt-0-6b-v2",
        "label": "Parakeet TDT 0.6B v2",
        "model": "Parakeet TDT 0.6B v2",
        "provider": "NVIDIA",
        "harness": "Open ASR published configuration",
        "effort": null,
        "score": 9.09,
        "costUsd": null,
        "uncertainty": null,
        "sourceUrl": "https://huggingface.co/datasets/hf-audio/open-asr-leaderboard-results/blob/ba5712d5ace8f785fa0daae1aecea8561ecd87c9/english_short_latest.csv",
        "details": [
          {
            "label": "Source model ID",
            "value": "nvidia/parakeet-tdt-0.6b-v2"
          },
          {
            "label": "Test",
            "value": "AMI-Cleaned · English · test split"
          },
          {
            "label": "Scoring",
            "value": "Publisher English normalization and compound-aware word-error rate"
          },
          {
            "label": "Exact execution configuration",
            "value": "Not published with this result; current launcher documentation is not a run receipt"
          },
          {
            "label": "Source snapshot",
            "value": "September 4, 2026; individual run date not published"
          }
        ]
      },
      {
        "id": "parakeet-tdt-0-6b-v3",
        "label": "Parakeet TDT 0.6B v3",
        "model": "Parakeet TDT 0.6B v3",
        "provider": "NVIDIA",
        "harness": "Open ASR published configuration",
        "effort": null,
        "score": 9.42,
        "costUsd": null,
        "uncertainty": null,
        "sourceUrl": "https://huggingface.co/datasets/hf-audio/open-asr-leaderboard-results/blob/ba5712d5ace8f785fa0daae1aecea8561ecd87c9/english_short_latest.csv",
        "details": [
          {
            "label": "Source model ID",
            "value": "nvidia/parakeet-tdt-0.6b-v3"
          },
          {
            "label": "Test",
            "value": "AMI-Cleaned · English · test split"
          },
          {
            "label": "Scoring",
            "value": "Publisher English normalization and compound-aware word-error rate"
          },
          {
            "label": "Exact execution configuration",
            "value": "Not published with this result; current launcher documentation is not a run receipt"
          },
          {
            "label": "Source snapshot",
            "value": "September 4, 2026; individual run date not published"
          }
        ]
      },
      {
        "id": "voxtral-small-24b-2507",
        "label": "Voxtral Small 24B 2507",
        "model": "Voxtral Small 24B 2507",
        "provider": "Mistral AI",
        "harness": "Open ASR published configuration",
        "effort": null,
        "score": 13.19,
        "costUsd": null,
        "uncertainty": null,
        "sourceUrl": "https://huggingface.co/datasets/hf-audio/open-asr-leaderboard-results/blob/ba5712d5ace8f785fa0daae1aecea8561ecd87c9/english_short_latest.csv",
        "details": [
          {
            "label": "Source model ID",
            "value": "mistralai/Voxtral-Small-24B-2507"
          },
          {
            "label": "Test",
            "value": "AMI-Cleaned · English · test split"
          },
          {
            "label": "Scoring",
            "value": "Publisher English normalization and compound-aware word-error rate"
          },
          {
            "label": "Exact execution configuration",
            "value": "Not published with this result; current launcher documentation is not a run receipt"
          },
          {
            "label": "Source snapshot",
            "value": "September 4, 2026; individual run date not published"
          }
        ]
      },
      {
        "id": "whisper-large-v3",
        "label": "Whisper large-v3",
        "model": "Whisper large-v3",
        "provider": "OpenAI",
        "harness": "Open ASR published configuration",
        "effort": null,
        "score": 13.63,
        "costUsd": null,
        "uncertainty": null,
        "sourceUrl": "https://huggingface.co/datasets/hf-audio/open-asr-leaderboard-results/blob/ba5712d5ace8f785fa0daae1aecea8561ecd87c9/english_short_latest.csv",
        "details": [
          {
            "label": "Source model ID",
            "value": "openai/whisper-large-v3"
          },
          {
            "label": "Test",
            "value": "AMI-Cleaned · English · test split"
          },
          {
            "label": "Scoring",
            "value": "Publisher English normalization and compound-aware word-error rate"
          },
          {
            "label": "Exact execution configuration",
            "value": "Not published with this result; current launcher documentation is not a run receipt"
          },
          {
            "label": "Source snapshot",
            "value": "September 4, 2026; individual run date not published"
          }
        ]
      },
      {
        "id": "omniasr-llm-7b-v2",
        "label": "omniASR LLM 7B v2",
        "model": "omniASR LLM 7B v2",
        "provider": "Meta",
        "harness": "Open ASR published configuration",
        "effort": null,
        "score": 13.86,
        "costUsd": null,
        "uncertainty": null,
        "sourceUrl": "https://huggingface.co/datasets/hf-audio/open-asr-leaderboard-results/blob/ba5712d5ace8f785fa0daae1aecea8561ecd87c9/english_short_latest.csv",
        "details": [
          {
            "label": "Source model ID",
            "value": "facebook/omniASR-LLM-7B-v2"
          },
          {
            "label": "Test",
            "value": "AMI-Cleaned · English · test split"
          },
          {
            "label": "Scoring",
            "value": "Publisher English normalization and compound-aware word-error rate"
          },
          {
            "label": "Exact execution configuration",
            "value": "Not published with this result; current launcher documentation is not a run receipt"
          },
          {
            "label": "Source snapshot",
            "value": "September 4, 2026; individual run date not published"
          }
        ]
      },
      {
        "id": "whisper-large-v3-turbo",
        "label": "Whisper large-v3-turbo",
        "model": "Whisper large-v3-turbo",
        "provider": "OpenAI",
        "harness": "Open ASR published configuration",
        "effort": null,
        "score": 13.88,
        "costUsd": null,
        "uncertainty": null,
        "sourceUrl": "https://huggingface.co/datasets/hf-audio/open-asr-leaderboard-results/blob/ba5712d5ace8f785fa0daae1aecea8561ecd87c9/english_short_latest.csv",
        "details": [
          {
            "label": "Source model ID",
            "value": "openai/whisper-large-v3-turbo"
          },
          {
            "label": "Test",
            "value": "AMI-Cleaned · English · test split"
          },
          {
            "label": "Scoring",
            "value": "Publisher English normalization and compound-aware word-error rate"
          },
          {
            "label": "Exact execution configuration",
            "value": "Not published with this result; current launcher documentation is not a run receipt"
          },
          {
            "label": "Source snapshot",
            "value": "September 4, 2026; individual run date not published"
          }
        ]
      }
    ]
  }
}