MODEL INTELLIGENCE
ModelRank
Compare models within one task and one source. Human preference, benchmark ability, and usage measure different things; their scores are not combined.
Methodology and data notes
Speech recognition. Audio generation is not covered here.
Metric: Average WER, lower is better
Ranking source
| Rank | Model | Organization | Average WER ↓ | 95% CI | Access | Evidence |
|---|---|---|---|---|---|---|
| 1 | Edge0/ARK-ASR-3B Average WER ↓ | Edge0 | 4.76 | — | Mixed / see source | Not team-checked Source |
| 2 | OpenMOSS-Team/MOSS-Transcribe-preview-2B Average WER ↓ | OpenMOSS | 4.87 | — | Mixed / see source | Not team-checked Source |
| 3 | OpenMOSS-Team/MOSS-Transcribe-Diarize Average WER ↓ | OpenMOSS | 5.17 | — | Mixed / see source | Not team-checked Source |
| 4 | hf-audio/MOSS-Transcribe-Diarize-HF Average WER ↓ | Hugging Face for Audio | 5.17 | — | Mixed / see source | Not team-checked Source |
| 5 | moondream/parakeet-ultra Average WER ↓ | moondream | 5.32 | — | Mixed / see source | Not team-checked Source |
| 6 | CohereLabs/cohere-transcribe-03-2026 Average WER ↓ | Cohere Labs | 5.42 | — | Mixed / see source | Not team-checked Source |
| 7 | Edge0/ARK-ASR-0.6B Average WER ↓ | Edge0 | 5.53 | — | Mixed / see source | Not team-checked Source |
| 8 | soundsgoodai/Zipformer-cr-ctc-transducer-XL-290M Average WER ↓ | SoundsGoodAI | 5.56 | — | Mixed / see source | Not team-checked Source |
| 9 | Qwen/Qwen3-ASR-1.7B Average WER ↓ | Qwen | 5.76 | — | Mixed / see source | Not team-checked Source |
| 10 | microsoft/Phi-4-multimodal-instruct Average WER ↓ | Microsoft | 6.02 | — | Mixed / see source | Not team-checked Source |
| 11 | nvidia/parakeet-tdt-0.6b-v2 Average WER ↓ | NVIDIA | 6.05 | — | Mixed / see source | Not team-checked Source |
| 12 | moondream/parakeet-redux Average WER ↓ | moondream | 6.05 | — | Mixed / see source | Not team-checked Source |
| 13 | LiquidAI/LFM3-Audio-2B Average WER ↓ | Liquid AI | 6.23 | — | Mixed / see source | Not team-checked Source |
| 14 | nvidia/canary-1b-flash Average WER ↓ | NVIDIA | 6.35 | — | Mixed / see source | Not team-checked Source |
| 15 | LiquidAI/LFM3-Omni-2B Average WER ↓ | Liquid AI | 6.36 | — | Mixed / see source | Not team-checked Source |
| 16 | kyutai/stt-2.6b-en Average WER ↓ | Kyutai | 6.40 | — | Mixed / see source | Not team-checked Source |
| 17 | Qwen/Qwen3-ASR-0.6B Average WER ↓ | Qwen | 6.42 | — | Mixed / see source | Not team-checked Source |
| 18 | nvidia/canary-1b Average WER ↓ | NVIDIA | 6.50 | — | Mixed / see source | Not team-checked Source |
| 19 | moonshine-ai/moonshine-streaming-medium Average WER ↓ | Moonshine AI | 6.66 | — | Mixed / see source | Not team-checked Source |
| 20 | soundsgoodai/Zipformer-transducer-XL-290M Average WER ↓ | SoundsGoodAI | 6.71 | — | Mixed / see source | Not team-checked Source |
| 21 | nvidia/parakeet-tdt-1.1b Average WER ↓ | NVIDIA | 7.01 | — | Mixed / see source | Not team-checked Source |
| 22 | zai-org/GLM-ASR-Nano-2512 Average WER ↓ | Z.ai | 7.03 | — | Mixed / see source | Not team-checked Source |
| 23 | Edge0/Audio8-ASR-0.1B Average WER ↓ | Edge0 | 7.03 | — | Mixed / see source | Not team-checked Source |
| 24 | mistralai/Voxtral-Mini-3B-2507 Average WER ↓ | Mistral | 7.05 | — | Mixed / see source | Not team-checked Source |
| 25 | nvidia/canary-180m-flash Average WER ↓ | NVIDIA | 7.12 | — | Mixed / see source | Not team-checked Source |
| 26 | nvidia/parakeet-rnnt-1.1b Average WER ↓ | NVIDIA | 7.12 | — | Mixed / see source | Not team-checked Source |
| 27 | nvidia/canary-1b-v2 Average WER ↓ | NVIDIA | 7.15 | — | Mixed / see source | Not team-checked Source |
| 28 | distil-whisper/distil-large-v3.5 Average WER ↓ | Whisper Distillation | 7.21 | — | Mixed / see source | Not team-checked Source |
| 29 | espnet/owsm_ctc_v4_1B Average WER ↓ | ESPnet | 7.42 | — | Mixed / see source | Not team-checked Source |
| 30 | openai/whisper-large-v3 Average WER ↓ | OpenAI | 7.44 | — | Mixed / see source | Not team-checked Source |
| 31 | nvidia/parakeet-tdt_ctc-110m Average WER ↓ | NVIDIA | 7.49 | — | Mixed / see source | Not team-checked Source |
| 32 | nvidia/parakeet-rnnt-0.6b Average WER ↓ | NVIDIA | 7.50 | — | Mixed / see source | Not team-checked Source |
| 33 | nvidia/parakeet-ctc-0.6b Average WER ↓ | NVIDIA | 7.69 | — | Mixed / see source | Not team-checked Source |
| 34 | microsoft/VibeVoice-ASR-HF Average WER ↓ | Microsoft | 7.77 | — | Mixed / see source | Not team-checked Source |
| 35 | openai/whisper-large-v2 Average WER ↓ | OpenAI | 7.83 | — | Mixed / see source | Not team-checked Source |
| 36 | moonshine-ai/moonshine-streaming-small Average WER ↓ | Moonshine AI | 7.84 | — | Mixed / see source | Not team-checked Source |
| 37 | openai/whisper-large Average WER ↓ | OpenAI | 7.94 | — | Mixed / see source | Not team-checked Source |
| 38 | openai/whisper-medium.en Average WER ↓ | OpenAI | 8.09 | — | Mixed / see source | Not team-checked Source |
| 39 | espnet/owsm_ctc_v3.1_1B Average WER ↓ | ESPnet | 8.12 | — | Mixed / see source | Not team-checked Source |
| 40 | nvidia/stt_en_conformer_ctc_large Average WER ↓ | NVIDIA | 8.32 | — | Mixed / see source | Not team-checked Source |
| 41 | speechbrain/asr-conformer-loquacious Average WER ↓ | SpeechBrain | 8.48 | — | Mixed / see source | Not team-checked Source |
| 42 | openai/whisper-small.en Average WER ↓ | OpenAI | 8.59 | — | Mixed / see source | Not team-checked Source |
| 43 | nvidia/stt_en_fastconformer_ctc_large Average WER ↓ | NVIDIA | 8.96 | — | Mixed / see source | Not team-checked Source |
| 44 | nvidia/stt_en_fastconformer_transducer_large Average WER ↓ | NVIDIA | 9.06 | — | Mixed / see source | Not team-checked Source |
| 45 | abr-ai/niagara-38m-batch.en Average WER ↓ | Applied Brain Research, Inc. | 9.36 | — | Mixed / see source | Not team-checked Source |
| 46 | moonshine-ai/moonshine-base Average WER ↓ | Moonshine AI | 9.99 | — | Mixed / see source | Not team-checked Source |
| 47 | openai/whisper-base.en Average WER ↓ | OpenAI | 10.32 | — | Mixed / see source | Not team-checked Source |
| 48 | abr-ai/niagara-19m-batch.en Average WER ↓ | Applied Brain Research, Inc. | 11.01 | — | Mixed / see source | Not team-checked Source |
| 49 | nvidia/stt_en_conformer_ctc_small Average WER ↓ | NVIDIA | 11.16 | — | Mixed / see source | Not team-checked Source |
| 50 | moonshine-ai/moonshine-streaming-tiny Average WER ↓ | Moonshine AI | 12 | — | Mixed / see source | Not team-checked Source |