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
Recorded meetings
Word error rate on AMI — real meetings recorded in a room, with people talking over each other and a microphone that is not close — LOWER is betterlower is betterProvenanceindependent leaderboard74 modelsscores Aug 2, 2026
1Modulate VfastModulate6.72——Aug 2, 2026leaderboard2Scribe v1Zoom6.94——Aug 2, 2026leaderboard3Azure Speech 06 2026Microsoft6.95——Aug 2, 2026leaderboard4Granite Speech 4.1 2b NARIBM2.3B6.96—~4.8Aug 2, 2026leaderboard5Cohere Transcribe 03 2026Cohere Labs2.1B7.01—~5.3Aug 2, 2026leaderboard6Granite Speech 4.1 2bIBM2.3B7.06—~4.9Aug 2, 2026leaderboard7Nyra Health CrisperWhisperNyra Health1.6B7.1—~7Aug 2, 2026leaderboard8Higgs Audio v3 STTBoson AI2.7B7.19—~4.2Aug 2, 2026leaderboard9Granite 4.0 1b SpeechIBM2.3B7.37—~5.1Aug 2, 2026leaderboard10Hojo ASR V1Hojo AI5.2B7.53—~2.3Aug 2, 2026leaderboard11Granite Speech 3.3 8bIBM8.6B7.7—~1.4Aug 2, 2026leaderboard12Granite Speech 3.3 2bIBM3B7.8—~4.1Aug 2, 2026leaderboard13MOSS Transcribe Preview 2BOpenMOSS2.4B7.8—~5.2Aug 2, 2026leaderboard14ARK ASR 3BAutoArk AI4.1B7.89—~3Aug 2, 2026leaderboard15Canary Qwen 2.5bNVIDIA2.6B7.91—~4.8Aug 2, 2026leaderboard16Scribe v2ElevenLabs8.2——Aug 2, 2026leaderboard17Resonant 1 FlashReson88.21——Aug 2, 2026leaderboard18Resonant 1Reson88.22——Aug 2, 2026leaderboard19Qwen3 ASR 1.7B HFQwen2B8.3—~6.4Aug 2, 2026leaderboard20Qwen3 ASR 1.7BQwen2B8.31—~6.4Aug 2, 2026leaderboard21MOSS Transcribe DiarizeOpenMOSS0.9B8.32—~15Aug 2, 2026leaderboard22Higgs Audio v3 8b STT v2Boson AI8.9B8.37—~1.5Aug 2, 2026leaderboard23ARK ASR 0.6BAutoArk AI1.3B8.6—~10Aug 2, 2026leaderboard24Nemotron Speech Streaming en 0.6bNVIDIA0.6B8.84—~23Aug 2, 2026leaderboard25Moonshine Streaming MediumUseful Sensors0.2B8.93—~60Aug 2, 2026leaderboard26Parakeet TDT 0.6b v2NVIDIA0.6B9.1—~24Aug 2, 2026leaderboard27Phi 4 Multimodal InstructMicrosoft5.6B9.11—~2.6Aug 2, 2026leaderboard28Smallest AI PulseSmallest AI9.12——Aug 2, 2026leaderboard29Qwen3 ASR 0.6B HFQwen0.8B9.32—~19Aug 2, 2026leaderboard30Parakeet TDT 0.6b v3NVIDIA0.6B9.41—~25Aug 2, 2026leaderboard31Qwen3 ASR 0.6BQwen0.8B9.47—~19Aug 2, 2026leaderboard32Avalon v1 enAqua Voice9.71——Aug 2, 2026leaderboard33Solaria 3Gladia9.89——Aug 2, 2026leaderboard34Speechmatics EnhancedSpeechmatics9.96——Aug 2, 2026leaderboard35Zipformer cr CTC Transducer XL 290MSounds Good AI0.3B10.28—~57Aug 2, 2026leaderboard36STT 2.6b enKyutai2.6B10.57—~6.4Aug 2, 2026leaderboard37Canary 1b FlashNVIDIA1B10.58—~17Aug 2, 2026leaderboard38Universal 3 5 ProAssemblyAI10.63——Aug 2, 2026leaderboard39Moonshine Streaming SmallUseful Sensors0.1B10.79—~180Aug 2, 2026leaderboard40Canary 1bNVIDIA1B10.87—~17Aug 2, 2026leaderboard41Audio8 ASR 0.1BAutoArk AI0.3B10.99—~58Aug 2, 2026leaderboard42Universal 3 ProAssemblyAI11.06——Aug 2, 2026leaderboard43Owsm CTC v4 1BESPnet1B11.11—~18Aug 2, 2026leaderboard44VibeVoice ASR HFMicrosoft8.3B12.03—~2.3Aug 2, 2026leaderboard45Distil Large v3.5Distil-Whisper0.8B12.09—~24Aug 2, 2026leaderboard46Canary 180m FlashNVIDIA0.2B12.1—~110Aug 2, 2026leaderboard47Parakeet CTC 1.1bNVIDIA1.1B12.24—~18Aug 2, 2026leaderboard48Parakeet TDT CTC 110mNVIDIA0.1B12.69—~180Aug 2, 2026leaderboard49Zipformer Transducer XL 290MSounds Good AI0.3B12.8—~71Aug 2, 2026leaderboard50Parakeet TDT 1.1bNVIDIA1.1B12.87—~19Aug 2, 2026leaderboard51Canary 1b v2NVIDIA1B13.02—~21Aug 2, 2026leaderboard52Parakeet CTC 0.6bNVIDIA0.6B13.1—~34Aug 2, 2026leaderboard53Voxtral Small 24B 2507Mistral AI24.3B13.1844~0.86Aug 2, 2026leaderboard54Voxtral Mini 4B Realtime 2602Mistral AI4.4B13.34—~4.8Aug 2, 2026leaderboard55Owsm CTC v3.1 1BESPnet1.1B13.35—~19Aug 2, 2026leaderboard56Nemotron 3.5 ASR Streaming 0.6bNVIDIA0.6B13.42—~34Aug 2, 2026leaderboard57Voxtral Mini 3B 2507Mistral AI5B13.56—~4.3Aug 2, 2026leaderboard58Whisper Large v3OpenAI1.5B13.63—~14Aug 2, 2026leaderboard59Lite Whisper Large v3 AccEfficient Speech1.4B13.69—~16Aug 2, 2026leaderboard60Niagara 38m Batch.enApplied Brain Research38M13.76—~690Aug 2, 2026leaderboard61omniASR LLM 7B v2Meta7.8B13.86——Aug 2, 2026leaderboard62Whisper Large v3 TurboOpenAI0.8B13.87—~28Aug 2, 2026leaderboard63Owsm CTC v3.2 ft 1BESPnet1B13.88—~22Aug 2, 2026leaderboard64GLM ASR Nano 2512Z.ai2.3B13.95—~9.6Aug 2, 2026leaderboard65Parakeet RNNT 1.1bNVIDIA1.1B14.02—~20Aug 2, 2026leaderboard66Parakeet RNNT 0.6bNVIDIA0.6B14.34—~38Aug 2, 2026leaderboard67ASR Conformer LargescaleasrSpeechBrain0.5B15.57—~52Aug 2, 2026leaderboard68Niagara 19m Batch.enApplied Brain Research20M16.71—~1700Aug 2, 2026leaderboard69Moonshine Streaming TinyUseful Sensors30M16.94—~850Aug 2, 2026leaderboard70omniASR CTC 7B v2Meta6.5B17.44——Aug 2, 2026leaderboard71Moonshine TinyUseful Sensors30M18.83—~940Aug 2, 2026leaderboard72Wav2vec2 Large 960h Lv60 SelfMeta0.3B25.83—~130Aug 2, 2026leaderboard73Hubert Large Ls960 ftMeta0.3B28.9—~140Aug 2, 2026leaderboard74MMS 1b AllMeta1B31.24—~50Aug 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.