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
Podcasts and video
Word error rate on GigaSpeech — podcasts, YouTube and audiobooks together, which is the closest set to everyday internet audio — LOWER is betterlower is betterProvenanceindependent leaderboard74 modelsscores Aug 2, 2026
1Hojo ASR V1Hojo AI5.2B6.14—~1.9Aug 2, 2026leaderboard2MOSS Transcribe Preview 2BOpenMOSS2.4B6.69—~4.4Aug 2, 2026leaderboard3ARK ASR 3BAutoArk AI4.1B6.94—~2.7Aug 2, 2026leaderboard4Higgs Audio v3 8b STT v2Boson AI8.9B7.09—~1.3Aug 2, 2026leaderboard5Qwen3 ASR 1.7BQwen2B7.22—~5.6Aug 2, 2026leaderboard6Qwen3 ASR 1.7B HFQwen2B7.23—~5.6Aug 2, 2026leaderboard7Modulate VfastModulate7.25——Aug 2, 2026leaderboard8Azure Speech 06 2026Microsoft7.26——Aug 2, 2026leaderboard9Solaria 3Gladia7.32——Aug 2, 2026leaderboard10ARK ASR 0.6BAutoArk AI1.3B7.38—~9Aug 2, 2026leaderboard11Resonant 1 FlashReson87.39——Aug 2, 2026leaderboard12Resonant 1Reson87.39——Aug 2, 2026leaderboard13Higgs Audio v3 STTBoson AI2.7B7.45—~4.4Aug 2, 2026leaderboard14Universal 3 ProAssemblyAI7.54——Aug 2, 2026leaderboard15Universal 3 5 ProAssemblyAI7.62——Aug 2, 2026leaderboard16Qwen3 ASR 0.6BQwen0.8B7.62—~16Aug 2, 2026leaderboard17Qwen3 ASR 0.6B HFQwen0.8B7.62—~15Aug 2, 2026leaderboard18Scribe v2ElevenLabs7.64——Aug 2, 2026leaderboard19MOSS Transcribe DiarizeOpenMOSS0.9B7.69—~13Aug 2, 2026leaderboard20Phi 4 Multimodal InstructMicrosoft5.6B7.8—~2.2Aug 2, 2026leaderboard21Canary Qwen 2.5bNVIDIA2.6B7.81—~4.8Aug 2, 2026leaderboard22Speechmatics EnhancedSpeechmatics7.84——Aug 2, 2026leaderboard23VibeVoice ASR HFMicrosoft8.3B7.87—~1.5Aug 2, 2026leaderboard24Scribe v1Zoom7.87——Aug 2, 2026leaderboard25Cohere Transcribe 03 2026Cohere Labs2.1B7.88—~6Aug 2, 2026leaderboard26Parakeet TDT 1.1bNVIDIA1.1B7.9—~11Aug 2, 2026leaderboard27Nemotron Speech Streaming en 0.6bNVIDIA0.6B7.91—~21Aug 2, 2026leaderboard28Parakeet TDT 0.6b v3NVIDIA0.6B8—~21Aug 2, 2026leaderboard29GLM ASR Nano 2512Z.ai2.3B8.02—~5.5Aug 2, 2026leaderboard30Avalon v1 enAqua Voice8.03——Aug 2, 2026leaderboard31Moonshine Streaming MediumUseful Sensors0.2B8.1—~54Aug 2, 2026leaderboard32Parakeet TDT 0.6b v2NVIDIA0.6B8.15—~21Aug 2, 2026leaderboard33Granite Speech 4.1 2bIBM2.3B8.18—~5.6Aug 2, 2026leaderboard34Smallest AI PulseSmallest AI8.18——Aug 2, 2026leaderboard35Canary 1b FlashNVIDIA1B8.19—~13Aug 2, 2026leaderboard36Distil Large v3.5Distil-Whisper0.8B8.21—~16Aug 2, 2026leaderboard37Parakeet RNNT 1.1bNVIDIA1.1B8.22—~12Aug 2, 2026leaderboard38STT 2.6b enKyutai2.6B8.22—~5Aug 2, 2026leaderboard39Zipformer cr CTC Transducer XL 290MSounds Good AI0.3B8.32—~46Aug 2, 2026leaderboard40Voxtral Small 24B 2507Mistral AI24.3B8.3328~0.54Aug 2, 2026leaderboard41Whisper Large v3OpenAI1.5B8.35—~8.8Aug 2, 2026leaderboard42Granite 4.0 1b SpeechIBM2.3B8.44—~5.8Aug 2, 2026leaderboard43Parakeet RNNT 0.6bNVIDIA0.6B8.45—~22Aug 2, 2026leaderboard44Canary 1bNVIDIA1B8.47—~13Aug 2, 2026leaderboard45Whisper Large v3 TurboOpenAI0.8B8.47—~17Aug 2, 2026leaderboard46Lite Whisper Large v3 AccEfficient Speech1.4B8.48—~9.6Aug 2, 2026leaderboard47Audio8 ASR 0.1BAutoArk AI0.3B8.48—~45Aug 2, 2026leaderboard48Granite Speech 3.3 8bIBM8.6B8.58—~1.6Aug 2, 2026leaderboard49Nyra Health CrisperWhisperNyra Health1.6B8.62—~8.5Aug 2, 2026leaderboard50Granite Speech 4.1 2b NARIBM2.3B8.62—~5.9Aug 2, 2026leaderboard51Voxtral Mini 3B 2507Mistral AI5B8.7—~2.8Aug 2, 2026leaderboard52Parakeet CTC 1.1bNVIDIA1.1B8.74—~13Aug 2, 2026leaderboard53Voxtral Mini 4B Realtime 2602Mistral AI4.4B8.8—~3.2Aug 2, 2026leaderboard54Parakeet TDT CTC 110mNVIDIA0.1B8.84—~130Aug 2, 2026leaderboard55Canary 180m FlashNVIDIA0.2B8.87—~81Aug 2, 2026leaderboard56Parakeet CTC 0.6bNVIDIA0.6B8.9—~23Aug 2, 2026leaderboard57Moonshine Streaming SmallUseful Sensors0.1B9.03—~150Aug 2, 2026leaderboard58Granite Speech 3.3 2bIBM3B9.03—~4.8Aug 2, 2026leaderboard59Canary 1b v2NVIDIA1B9.12—~14Aug 2, 2026leaderboard60omniASR LLM 7B v2Meta7.8B9.21——Aug 2, 2026leaderboard61Zipformer Transducer XL 290MSounds Good AI0.3B9.21—~51Aug 2, 2026leaderboard62Owsm CTC v4 1BESPnet1B9.29—~15Aug 2, 2026leaderboard63ASR Conformer LargescaleasrSpeechBrain0.5B9.7—~32Aug 2, 2026leaderboard64Nemotron 3.5 ASR Streaming 0.6bNVIDIA0.6B9.86—~25Aug 2, 2026leaderboard65Owsm CTC v3.2 ft 1BESPnet1B10.14—~16Aug 2, 2026leaderboard66Owsm CTC v3.1 1BESPnet1.1B10.35—~15Aug 2, 2026leaderboard67Niagara 38m Batch.enApplied Brain Research38M11.31—~570Aug 2, 2026leaderboard68Moonshine Streaming TinyUseful Sensors30M12.46—~620Aug 2, 2026leaderboard69Moonshine TinyUseful Sensors30M12.6—~630Aug 2, 2026leaderboard70omniASR CTC 7B v2Meta6.5B12.95——Aug 2, 2026leaderboard71Niagara 19m Batch.enApplied Brain Research20M14.07—~1400Aug 2, 2026leaderboard72Wav2vec2 Large 960h Lv60 SelfMeta0.3B14.76—~74Aug 2, 2026leaderboard73Hubert Large Ls960 ftMeta0.3B16.02—~80Aug 2, 2026leaderboard74MMS 1b AllMeta1B17.15—~27Aug 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.