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
Financial calls
Word error rate on SPGISpeech — corporate earnings calls with professional reference transcripts — LOWER is betterlower is betterProvenanceindependent leaderboard74 modelsscores Aug 2, 2026
1Scribe v1Zoom1.43——Aug 2, 2026leaderboard2Universal 3 ProAssemblyAI1.52——Aug 2, 2026leaderboard3MOSS Transcribe Preview 2BOpenMOSS2.4B1.61—~1.1Aug 2, 2026leaderboard4Zipformer cr CTC Transducer XL 290MSounds Good AI0.3B1.64—~9.1Aug 2, 2026leaderboard5Canary Qwen 2.5bNVIDIA2.6B1.7—~1Aug 2, 2026leaderboard6MOSS Transcribe DiarizeOpenMOSS0.9B1.7—~3Aug 2, 2026leaderboard7Hojo ASR V1Hojo AI5.2B1.74—~0.53Aug 2, 2026leaderboard8Canary 1b FlashNVIDIA1B1.75—~2.8Aug 2, 2026leaderboard9Universal 3 5 ProAssemblyAI1.81——Aug 2, 2026leaderboard10Voxtral Small 24B 2507Mistral AI24.3B1.846~0.12Aug 2, 2026leaderboard11GLM ASR Nano 2512Z.ai2.3B1.86—~1.3Aug 2, 2026leaderboard12Canary 1bNVIDIA1B1.86—~3Aug 2, 2026leaderboard13STT 2.6b enKyutai2.6B1.87—~1.1Aug 2, 2026leaderboard14Avalon v1 enAqua Voice1.88——Aug 2, 2026leaderboard15Zipformer Transducer XL 290MSounds Good AI0.3B1.91—~11Aug 2, 2026leaderboard16Smallest AI PulseSmallest AI1.91——Aug 2, 2026leaderboard17Nyra Health CrisperWhisperNyra Health1.6B1.94—~1.9Aug 2, 2026leaderboard18Parakeet TDT 0.6b v2NVIDIA0.6B1.94—~5.1Aug 2, 2026leaderboard19Canary 1b v2NVIDIA1B2.03—~3.2Aug 2, 2026leaderboard20Canary 180m FlashNVIDIA0.2B2.04—~19Aug 2, 2026leaderboard21Higgs Audio v3 STTBoson AI2.7B2.07—~1.2Aug 2, 2026leaderboard22Voxtral Mini 3B 2507Mistral AI5B2.15—~0.68Aug 2, 2026leaderboard23Voxtral Mini 4B Realtime 2602Mistral AI4.4B2.23—~0.81Aug 2, 2026leaderboard24Parakeet TDT CTC 110mNVIDIA0.1B2.26—~32Aug 2, 2026leaderboard25Moonshine Streaming MediumUseful Sensors0.2B2.26—~15Aug 2, 2026leaderboard26Azure Speech 06 2026Microsoft2.31——Aug 2, 2026leaderboard27Owsm CTC v4 1BESPnet1B2.31—~3.7Aug 2, 2026leaderboard28Modulate VfastModulate2.41——Aug 2, 2026leaderboard29ARK ASR 3BAutoArk AI4.1B2.46—~0.95Aug 2, 2026leaderboard30Owsm CTC v3.2 ft 1BESPnet1B2.47—~3.9Aug 2, 2026leaderboard31Parakeet RNNT 1.1bNVIDIA1.1B2.56—~3.7Aug 2, 2026leaderboard32Qwen3 ASR 1.7B HFQwen2B2.57—~2Aug 2, 2026leaderboard33Qwen3 ASR 1.7BQwen2B2.58—~2Aug 2, 2026leaderboard34Distil Large v3.5Distil-Whisper0.8B2.62—~5.2Aug 2, 2026leaderboard35Lite Whisper Large v3 AccEfficient Speech1.4B2.62—~3Aug 2, 2026leaderboard36ARK ASR 0.6BAutoArk AI1.3B2.63—~3.2Aug 2, 2026leaderboard37Owsm CTC v3.1 1BESPnet1.1B2.64—~3.8Aug 2, 2026leaderboard38Solaria 3Gladia2.68——Aug 2, 2026leaderboard39Nemotron Speech Streaming en 0.6bNVIDIA0.6B2.69—~7.1Aug 2, 2026leaderboard40Whisper Large v3OpenAI1.5B2.71—~2.9Aug 2, 2026leaderboard41Parakeet TDT 1.1bNVIDIA1.1B2.72—~3.9Aug 2, 2026leaderboard42Phi 4 Multimodal InstructMicrosoft5.6B2.73—~0.77Aug 2, 2026leaderboard43Qwen3 ASR 0.6BQwen0.8B2.74—~5.6Aug 2, 2026leaderboard44Qwen3 ASR 0.6B HFQwen0.8B2.74—~5.5Aug 2, 2026leaderboard45Cohere Transcribe 03 2026Cohere Labs2.1B2.74—~2.1Aug 2, 2026leaderboard46Resonant 1Reson82.77——Aug 2, 2026leaderboard47Whisper Large v3 TurboOpenAI0.8B2.79—~5.6Aug 2, 2026leaderboard48Resonant 1 FlashReson82.8——Aug 2, 2026leaderboard49Speechmatics EnhancedSpeechmatics2.81——Aug 2, 2026leaderboard50Moonshine Streaming SmallUseful Sensors0.1B2.85—~48Aug 2, 2026leaderboard51Parakeet RNNT 0.6bNVIDIA0.6B2.88—~7.6Aug 2, 2026leaderboard52Scribe v2ElevenLabs3.01——Aug 2, 2026leaderboard53Niagara 38m Batch.enApplied Brain Research38M3.08—~150Aug 2, 2026leaderboard54omniASR LLM 7B v2Meta7.8B3.22——Aug 2, 2026leaderboard55Higgs Audio v3 8b STT v2Boson AI8.9B3.23—~0.58Aug 2, 2026leaderboard56Granite Speech 4.1 2b NARIBM2.3B3.23—~2.2Aug 2, 2026leaderboard57Nemotron 3.5 ASR Streaming 0.6bNVIDIA0.6B3.27—~8.2Aug 2, 2026leaderboard58Granite Speech 4.1 2bIBM2.3B3.47—~2.4Aug 2, 2026leaderboard59VibeVoice ASR HFMicrosoft8.3B3.47—~0.66Aug 2, 2026leaderboard60Parakeet CTC 0.6bNVIDIA0.6B3.47—~9.1Aug 2, 2026leaderboard61Granite Speech 3.3 2bIBM3B3.5—~1.9Aug 2, 2026leaderboard62Parakeet CTC 1.1bNVIDIA1.1B3.53—~5.1Aug 2, 2026leaderboard63Granite Speech 3.3 8bIBM8.6B3.54—~0.65Aug 2, 2026leaderboard64Granite 4.0 1b SpeechIBM2.3B3.55—~2.4Aug 2, 2026leaderboard65Parakeet TDT 0.6b v3NVIDIA0.6B3.63—~9.6Aug 2, 2026leaderboard66ASR Conformer LargescaleasrSpeechBrain0.5B3.69—~12Aug 2, 2026leaderboard67Audio8 ASR 0.1BAutoArk AI0.3B3.73—~20Aug 2, 2026leaderboard68Niagara 19m Batch.enApplied Brain Research20M3.86—~390Aug 2, 2026leaderboard69omniASR CTC 7B v2Meta6.5B5.16——Aug 2, 2026leaderboard70Moonshine Streaming TinyUseful Sensors30M5.71—~290Aug 2, 2026leaderboard71Moonshine TinyUseful Sensors30M6.88—~340Aug 2, 2026leaderboard72MMS 1b AllMeta1B7.49—~12Aug 2, 2026leaderboard73Wav2vec2 Large 960h Lv60 SelfMeta0.3B7.81—~39Aug 2, 2026leaderboard74Hubert Large Ls960 ftMeta0.3B8.95—~45Aug 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.