Leaderboards
Benchmarks
Every score carries its origin — leaderboard is independent,vendor is self-reported. History is kept, never overwritten.
Accented speech74Clean read speech74European-accented speech74Financial calls74Harder read speech74Open ASR RTFx62Open ASR WER74Podcasts and video74Recorded meetings74
Harder read speech
Word error rate on LibriSpeech test-other — the same reading task on noisier, more accented recordings — LOWER is betterlower is betterProvenanceindependent leaderboard74 modelsscores Aug 2, 2026
1Modulate VfastModulate1.88——Aug 2, 2026leaderboard2Cohere Transcribe 03 2026Cohere Labs2.1B2.04—~1.5Aug 2, 2026leaderboard3Higgs Audio v3 8b STT v2Boson AI8.9B2.05—~0.37Aug 2, 2026leaderboard4Resonant 1Reson82.08——Aug 2, 2026leaderboard5Scribe v1Zoom2.08——Aug 2, 2026leaderboard6Universal 3 5 ProAssemblyAI2.1——Aug 2, 2026leaderboard7Parakeet RNNT 1.1bNVIDIA1.1B2.1—~3Aug 2, 2026leaderboard8Resonant 1 FlashReson82.13——Aug 2, 2026leaderboard9Granite Speech 4.1 2bIBM2.3B2.17—~1.5Aug 2, 2026leaderboard10Parakeet TDT 1.1bNVIDIA1.1B2.22—~3.2Aug 2, 2026leaderboard11Universal 3 ProAssemblyAI2.27——Aug 2, 2026leaderboard12Scribe v2ElevenLabs2.32——Aug 2, 2026leaderboard13ARK ASR 3BAutoArk AI4.1B2.35—~0.91Aug 2, 2026leaderboard14Granite Speech 4.1 2b NARIBM2.3B2.4—~1.7Aug 2, 2026leaderboard15Granite 4.0 1b SpeechIBM2.3B2.49—~1.7Aug 2, 2026leaderboard16Azure Speech 06 2026Microsoft2.49——Aug 2, 2026leaderboard17Granite Speech 3.3 8bIBM8.6B2.52—~0.46Aug 2, 2026leaderboard18Canary 1b FlashNVIDIA1B2.52—~4Aug 2, 2026leaderboard19Canary 1bNVIDIA1B2.53—~4Aug 2, 2026leaderboard20Higgs Audio v3 STTBoson AI2.7B2.6—~1.5Aug 2, 2026leaderboard21Canary Qwen 2.5bNVIDIA2.6B2.62—~1.6Aug 2, 2026leaderboard22Parakeet RNNT 0.6bNVIDIA0.6B2.64—~6.9Aug 2, 2026leaderboard23Smallest AI PulseSmallest AI2.7——Aug 2, 2026leaderboard24Parakeet TDT 0.6b v2NVIDIA0.6B2.73—~7.2Aug 2, 2026leaderboard25Voxtral Small 24B 2507Mistral AI24.3B2.799~0.18Aug 2, 2026leaderboard26MOSS Transcribe Preview 2BOpenMOSS2.4B2.8—~1.9Aug 2, 2026leaderboard27Granite Speech 3.3 2bIBM3B2.84—~1.5Aug 2, 2026leaderboard28Avalon v1 enAqua Voice2.87——Aug 2, 2026leaderboard29Qwen3 ASR 1.7BQwen2B2.92—~2.3Aug 2, 2026leaderboard30Qwen3 ASR 1.7B HFQwen2B2.93—~2.3Aug 2, 2026leaderboard31Solaria 3Gladia2.98——Aug 2, 2026leaderboard32Zipformer cr CTC Transducer XL 290MSounds Good AI0.3B3.01—~17Aug 2, 2026leaderboard33Canary 1b v2NVIDIA1B3.07—~4.9Aug 2, 2026leaderboard34Parakeet TDT 0.6b v3NVIDIA0.6B3.12—~8.2Aug 2, 2026leaderboard35Parakeet CTC 1.1bNVIDIA1.1B3.23—~4.7Aug 2, 2026leaderboard36Hojo ASR V1Hojo AI5.2B3.24—~0.99Aug 2, 2026leaderboard37omniASR LLM 7B v2Meta7.8B3.25——Aug 2, 2026leaderboard38Wav2vec2 Large 960h Lv60 SelfMeta0.3B3.33—~17Aug 2, 2026leaderboard39Canary 180m FlashNVIDIA0.2B3.41—~31Aug 2, 2026leaderboard40Parakeet CTC 0.6bNVIDIA0.6B3.41—~9Aug 2, 2026leaderboard41ARK ASR 0.6BAutoArk AI1.3B3.44—~4.2Aug 2, 2026leaderboard42Phi 4 Multimodal InstructMicrosoft5.6B3.45—~0.98Aug 2, 2026leaderboard43Lite Whisper Large v3 AccEfficient Speech1.4B3.5—~4Aug 2, 2026leaderboard44Whisper Large v3OpenAI1.5B3.52—~3.7Aug 2, 2026leaderboard45Voxtral Mini 3B 2507Mistral AI5B3.62—~1.1Aug 2, 2026leaderboard46Hubert Large Ls960 ftMeta0.3B3.65—~18Aug 2, 2026leaderboard47Whisper Large v3 TurboOpenAI0.8B3.7—~7.4Aug 2, 2026leaderboard48GLM ASR Nano 2512Z.ai2.3B3.82—~2.6Aug 2, 2026leaderboard49Nyra Health CrisperWhisperNyra Health1.6B3.95—~3.9Aug 2, 2026leaderboard50Qwen3 ASR 0.6BQwen0.8B3.97—~8.1Aug 2, 2026leaderboard51STT 2.6b enKyutai2.6B3.99—~2.4Aug 2, 2026leaderboard52Qwen3 ASR 0.6B HFQwen0.8B4.01—~8Aug 2, 2026leaderboard53ASR Conformer LargescaleasrSpeechBrain0.5B4.12—~14Aug 2, 2026leaderboard54MOSS Transcribe DiarizeOpenMOSS0.9B4.14—~7.3Aug 2, 2026leaderboard55Speechmatics EnhancedSpeechmatics4.26——Aug 2, 2026leaderboard56Nemotron Speech Streaming en 0.6bNVIDIA0.6B4.34—~11Aug 2, 2026leaderboard57Owsm CTC v4 1BESPnet1B4.37—~6.9Aug 2, 2026leaderboard58Distil Large v3.5Distil-Whisper0.8B4.5—~9Aug 2, 2026leaderboard59Zipformer Transducer XL 290MSounds Good AI0.3B4.54—~25Aug 2, 2026leaderboard60Moonshine Streaming MediumUseful Sensors0.2B4.6—~31Aug 2, 2026leaderboard61Owsm CTC v3.1 1BESPnet1.1B4.65—~6.7Aug 2, 2026leaderboard62Parakeet TDT CTC 110mNVIDIA0.1B4.71—~67Aug 2, 2026leaderboard63Owsm CTC v3.2 ft 1BESPnet1B4.82—~7.7Aug 2, 2026leaderboard64Voxtral Mini 4B Realtime 2602Mistral AI4.4B4.92—~1.8Aug 2, 2026leaderboard65VibeVoice ASR HFMicrosoft8.3B4.93—~0.94Aug 2, 2026leaderboard66omniASR CTC 7B v2Meta6.5B5.05——Aug 2, 2026leaderboard67Moonshine Streaming SmallUseful Sensors0.1B6.37—~110Aug 2, 2026leaderboard68Audio8 ASR 0.1BAutoArk AI0.3B6.59—~35Aug 2, 2026leaderboard69Nemotron 3.5 ASR Streaming 0.6bNVIDIA0.6B6.78—~17Aug 2, 2026leaderboard70MMS 1b AllMeta1B7.2—~11Aug 2, 2026leaderboard71Niagara 38m Batch.enApplied Brain Research38M9.22—~460Aug 2, 2026leaderboard72Niagara 19m Batch.enApplied Brain Research20M10.95—~1100Aug 2, 2026leaderboard73Moonshine TinyUseful Sensors30M11.06—~550Aug 2, 2026leaderboard74Moonshine Streaming TinyUseful Sensors30M11.54—~580Aug 2, 2026leaderboard
A tilde in Score / GB VRAM means the memory figure behind it is calculated from the parameter count rather than read from a published file. Those ratios are shown to two significant figures, and rows within a few percent of each other are not distinguishable.