whatmodel.sh · find the best speech-to-text models you can run

Best speech-to-text models to run locally on a 12GB graphics card

A leaderboard of the best speech-to-text models you can run locally on a 12GB graphics card (with 32GB of system RAM). Granite Speech 5.0 470M TurboCTC NC is #1 right now, followed by Cohere Transcribe Arabic (07-2026) and Granite Speech 4.1 2B. Models are ranked by one score that combines 86 published benchmarks. Each model shows the download that fits and the command to run it.

130 models · scores combine 86 benchmarks · updated how it works · find a leaderboard · benchmark your machine
task[all tasks][long-form transcription][multilingual transcription]
Models for 32GB RAM · 12GB graphics card → ~12GB fast · local only · reset
+ more
card[RTX 2060 12GB][RTX 3060 12GB][RTX 3080 12GB][RTX 3080 Ti][RTX 4070][RTX 4070 GDDR6][RTX 4070 Super][RTX 4070 Ti][RTX 5070][RTX 4080 Laptop][RTX 5070 Laptop][RTX 5070 Ti Laptop]
set[all sets][Open ASR English short-form WER (March 2026) · 176][LibriSpeech test-clean WER · 2]

Filter the model list while keeping overall scores. Models without a shared score scale are not yet calibrated; select a set to see its own scores.

Detect your exact chip or graphics card:

curl -G whatmodel.sh/stt -d max=0 --data-urlencode "mac=$(sysctl -n hw.memsize machdep.cpu.brand_string)"Mac

curl -G whatmodel.sh/stt -d max=0 --data-urlencode "ram=$(awk '/MemT/{print $2}' /proc/meminfo)" --data-urlencode "nv=$(nvidia-smi --query-gpu=name,memory.total --format=csv,noheader 2>/dev/null)"Linux

curl.exe -G whatmodel.sh/stt -d max=0 --data-urlencode "ram=$((gcim Win32_ComputerSystem).TotalPhysicalMemory)" --data-urlencode "nv=$(nvidia-smi --query-gpu=name,memory.total --format=csv,noheader 2>$null)"Windows

$ terminal

detecting your OS…

Run curl "whatmodel.sh/stt?ram=32&gpu=card&vram=12&max=0"to get this list in your terminal (add &format=json for agents).

ram[8][16][24][32][48][64][96][128][192][256]GB
gpu[apple silicon][graphics card][none]
vram[1][2][4][8][12][16][24][32][48][64][80][96][128][160]GB
price[local][$0.001][$0.002][$0.005][$0.01][any]max per audio min

detecting your OS…

Score vs memory to run

#modelscore↓where↓speed↓RAM↓get itsources
#modelscore↓where↓speed↓RAM↓get itsources
15Granite Speech 5.0 470M TurboCTC NC · q8_0-gguf (transcribe.cpp) IBM64±4*local~0.5GBhow to run › transcribe.cpp model catalog · LibriSpeech test-clean WER of validated full-precision ports
16Cohere Transcribe Arabic (07-2026) · bf16 (transformers >=5.4.0) Cohere63±3local~4.1GBhow to run › transcribe.cpp model catalog · LibriSpeech test-clean WER of validated full-precision portsQCRI Arabic + English ASR leaderboard (Hugging Face Space QCRI/AURA-Arabic-English-ASR-leaderboard)Open Universal Arabic ASR Leaderboard (Elm Research Center)
18Granite Speech 4.1 2B · bf16 IBM62±3local~6GBhow to run › transcribe.cpp model catalog · LibriSpeech test-clean WER of validated full-precision portsSwitch-aware evaluation of ASR on English-Yoruba code-switched speech (arXiv:2609.11786v1)
19Granite 4.0 1B Speech · bf16 IBM62±3local~6GBhow to run › transcribe.cpp model catalog · LibriSpeech test-clean WER of validated full-precision portsOpen ASR Leaderboard
25Cohere Transcribe (03-2026) · bf16 Cohere62±2local~5.5GBhow to run › Open ASR Leaderboardtranscribe.cpp model catalog · LibriSpeech test-clean WER of validated full-precision portsDutch ASR Leaderboard (tvosch) · committed result filesOpen Universal Arabic ASR Leaderboard (Elm Research Center)Quranic ASR Leaderboard (Quran-Lab, held-out recitation set)Dysarthric ASR with clinical context on the Speech Accessibility Project (arXiv:2605.02782v2)
26Granite Speech 4.1 2B NAR · q8_0-gguf (transcribe.cpp) IBM62±4*local~2.5GBhow to run › transcribe.cpp model catalog · LibriSpeech test-clean WER of validated full-precision ports
28Canary-Qwen-2.5B · bf16 NVIDIA61±2local~6.6GBhow to run › NVIDIA ASR model cards (Parakeet / Canary, Hugging Face)AfriSpeech-MultiBench (arXiv:2511.14255v1) · African-accented English ASR benchmark suiteOpen ASR Leaderboardtranscribe.cpp model catalog · LibriSpeech test-clean WER of validated full-precision portsRW-Voice-EQ Bench (Hume AI)Switch-aware evaluation of ASR on English-Yoruba code-switched speech (arXiv:2609.11786v1)
34Granite Speech 4.1 2B Plus · q8_0-gguf (transcribe.cpp) IBM61±4*local~2.4GBhow to run › transcribe.cpp model catalog · LibriSpeech test-clean WER of validated full-precision ports
35Parakeet TDT 0.6B v2 · fp32 NVIDIA60±1local~3.4GBhow to run › NVIDIA ASR model cards (Parakeet / Canary, Hugging Face)AfriSpeech-MultiBench (arXiv:2511.14255v1) · African-accented English ASR benchmark suiteOpen ASR Leaderboardtranscribe.cpp model catalog · LibriSpeech test-clean WER of validated full-precision portsmicrosoft/VibeVoice-ASR-BitNet model card (Accuracy table)Switch-aware evaluation of ASR on English-Yoruba code-switched speech (arXiv:2609.11786v1)+1
37Parakeet Unified EN 0.6B · q8_0-gguf (transcribe.cpp) NVIDIA60±4*local~0.7GBhow to run › transcribe.cpp model catalog · LibriSpeech test-clean WER of validated full-precision ports
39Qwen3-ASR-1.7B · bf16 Alibaba (Qwen)60±1local~6.1GBhow to run › Qwen/Qwen3-ASR-1.7B model cardOpen ASR Leaderboardtranscribe.cpp model catalog · LibriSpeech test-clean WER of validated full-precision portsGigaSpeechBench: A Real-World Multilingual Speech-to-Text BenchmarkMega-ASR (arXiv:2605.19833v1) · adverse-condition ASR baselines (CHiME-4, VOiCES, NOIZEUS)RW-Voice-EQ Bench (Hume AI)+10
40Canary-1B-Flash · fp32 NVIDIA60±2local~4.4GBhow to run › AfriSpeech-MultiBench (arXiv:2511.14255v1) · African-accented English ASR benchmark suitetranscribe.cpp model catalog · LibriSpeech test-clean WER of validated full-precision portsOpen ASR Leaderboard
41Granite Speech 3.3 2B · bf16 IBM60±3local~7.7GBhow to run › AfriSpeech-MultiBench (arXiv:2511.14255v1) · African-accented English ASR benchmark suiteOpen ASR Leaderboard
45Granite Speech 3.3 8B · bf16 IBM59±5*local (slow)~21GBhow to run › Open ASR LeaderboardDysarthric ASR with clinical context on the Speech Accessibility Project (arXiv:2605.02782v2)
49Canary-1B · fp32 NVIDIA59±3local~5.3GBhow to run › transcribe.cpp model catalog · LibriSpeech test-clean WER of validated full-precision portsOpen ASR LeaderboardSpeech Robust Bench (arXiv:2403.07937v3) · clean-speech WER of 25+ open ASR models
50Parakeet Ultra · q8_0-gguf (transcribe.cpp) Moondream59±3local~0.7GBhow to run › Moondream parakeet-ultra and parakeet-redux model cards (Benchmarks sections)transcribe.cpp model catalog · LibriSpeech test-clean WER of validated full-precision ports
51Parakeet TDT 1.1B · fp32 NVIDIA59±2local~5.8GBhow to run › AfriSpeech-MultiBench (arXiv:2511.14255v1) · African-accented English ASR benchmark suiteOpen ASR Leaderboardtranscribe.cpp model catalog · LibriSpeech test-clean WER of validated full-precision portsDysarthric ASR with clinical context on the Speech Accessibility Project (arXiv:2605.02782v2)
52Voxtral Mini 3B (2507) · bf16 Mistral58±1local (slow)~12GBhow to run › Mistral AI — Voxtral technical reportOpen ASR LeaderboardAfriSpeech-MultiBench (arXiv:2511.14255v1) · African-accented English ASR benchmark suitetranscribe.cpp model catalog · LibriSpeech test-clean WER of validated full-precision portsMega-ASR (arXiv:2605.19833v1) · adverse-condition ASR baselines (CHiME-4, VOiCES, NOIZEUS)Open ASR Leaderboard – Chilean Spanish (idsudd / Alonso Astroza)+5
54Parakeet RNNT 1.1B · fp32 NVIDIA58±2local~5.6GBhow to run › NVIDIA ASR model cards (Parakeet / Canary, Hugging Face)AfriSpeech-MultiBench (arXiv:2511.14255v1) · African-accented English ASR benchmark suiteOpen ASR Leaderboardtranscribe.cpp model catalog · LibriSpeech test-clean WER of validated full-precision ports
56Canary-1B-v2 · onnx-fp32 NVIDIA58±1local~5.2GBhow to run › NVIDIA ASR model cards (Parakeet / Canary, Hugging Face)Open ASR Leaderboardtranscribe.cpp model catalog · LibriSpeech test-clean WER of validated full-precision portsCanary-1B-v2 & Parakeet-TDT-0.6B-v3: Efficient and High-Performance Models for Multilingual ASR and AST (NVIDIA, arXiv:2509.14128v2)Mega-ASR (arXiv:2605.19833v1) · adverse-condition ASR baselines (CHiME-4, VOiCES, NOIZEUS)Open ASR Leaderboard – Chilean Spanish (idsudd / Alonso Astroza)+2
57Phi-4-multimodal-instruct · bf16 Microsoft58±2local (slow)~14GBhow to run › Open ASR LeaderboardAfriSpeech-MultiBench (arXiv:2511.14255v1) · African-accented English ASR benchmark suiteOpen ASR Leaderboard – Chilean Spanish (idsudd / Alonso Astroza)Benchmarking Automatic Speech Recognition Tools for Iberian Languages (arXiv:2609.36920v1), Table 3Dysarthric ASR with clinical context on the Speech Accessibility Project (arXiv:2605.02782v2)
58Omnilingual ASR LLM 7B · bf16 Meta58±2local (slow)~19GBhow to run › Open ASR LeaderboardOmnilingual ASR: Open-Source Multilingual Speech Recognition for 1600+ Languages (arXiv:2511.09690v1)Open ASR Leaderboard – Chilean Spanish (idsudd / Alonso Astroza)Open Universal Arabic ASR Leaderboard (Elm Research Center)Quranic ASR Leaderboard (Quran-Lab, held-out recitation set)
59Omnilingual ASR LLM 7B v2 · bf16 Meta58±3*local (slow)~19GBhow to run › Open ASR Leaderboard
60Distil-Whisper distil-large-v3.5 · fp16 Hugging Face (Distil-Whisper)58±1local~2.3GBhow to run › distil-whisper/distil-large-v3.5 model cardAfriSpeech-MultiBench (arXiv:2511.14255v1) · African-accented English ASR benchmark suiteOpen ASR LeaderboardRW-Voice-EQ Bench (Hume AI)
61Whisper large-v3-turbo · fp16 OpenAI57±1local~2.4GBhow to run › distil-whisper/distil-large-v3.5 model cardOpen ASR Leaderboardtranscribe.cpp model catalog · LibriSpeech test-clean WER of validated full-precision portsverliapp/stt-benchmark · independent STT accuracy benchmarkLiteASR (arXiv:2502.20583v2) · ESB WER of Whisper and LiteASR-compressed encodersRussian ASR Leaderboard (ESpeech; benchmark by Nikolay Shmyrev)+7
62Omnilingual ASR LLM 3B · bf16 Meta57±2local~11GBhow to run › Open ASR LeaderboardOmnilingual ASR: Open-Source Multilingual Speech Recognition for 1600+ Languages (arXiv:2511.09690v1)Open Universal Arabic ASR Leaderboard (Elm Research Center)Quranic ASR Leaderboard (Quran-Lab, held-out recitation set)
63Fun-ASR-MLT-Nano-2512 · q8_0-gguf (transcribe.cpp) Alibaba (FunAudioLLM)57±4local~0.9GBhow to run › transcribe.cpp model catalog · LibriSpeech test-clean WER of validated full-precision ports
64Parakeet Redux · ternary 1.58-bit (Photon) Moondream57±4†local~0.2GBhow to run › Moondream parakeet-ultra and parakeet-redux model cards (Benchmarks sections)
65Parakeet TDT 0.6B v3 · fp32 NVIDIA57±1local~3.5GBhow to run › NVIDIA ASR model cards (Parakeet / Canary, Hugging Face)Open ASR LeaderboardMoondream parakeet-ultra and parakeet-redux model cards (Benchmarks sections)transcribe.cpp model catalog · LibriSpeech test-clean WER of validated full-precision portsCanary-1B-v2 & Parakeet-TDT-0.6B-v3: Efficient and High-Performance Models for Multilingual ASR and AST (NVIDIA, arXiv:2509.14128v2)Russian STT benchmark by domain+11
66Kyutai STT 2.6B (en) · bf16 Kyutai57±4local~6.7GBhow to run › Voxtral RealtimeOpen ASR Leaderboard
67Whisper large-v3 · fp16 OpenAI57±1local~4.2GBhow to run › Mistral AI — Voxtral technical reportOpen ASR LeaderboardQwen/Qwen3-ASR-1.7B model carddistil-whisper/distil-large-v3.5 model cardAfriSpeech-MultiBench (arXiv:2511.14255v1) · African-accented English ASR benchmark suitetranscribe.cpp model catalog · LibriSpeech test-clean WER of validated full-precision ports+25
68GLM-ASR-Nano-2512 · bf16 Z.ai57±4local~5.9GBhow to run › Open ASR Leaderboardzai-org/GLM-ASR README benchmark table (GLM-ASR-Nano vs open/closed ASR)
69Parakeet RNNT 0.6B · fp32 NVIDIA57±3local~3.5GBhow to run › Open ASR Leaderboardtranscribe.cpp model catalog · LibriSpeech test-clean WER of validated full-precision ports
71Orukeet · q8_0-gguf (transcribe.cpp) Oruk AI57±4*local~0.7GBhow to run › transcribe.cpp model catalog · LibriSpeech test-clean WER of validated full-precision ports
73Parakeet TDT-CTC 1.1B · q8_0-gguf (transcribe.cpp) NVIDIA56±4*local~1.3GBhow to run › transcribe.cpp model catalog · LibriSpeech test-clean WER of validated full-precision ports
75Omnilingual ASR LLM 1B v2 · bf16 Meta56±3*local~6GBhow to run › Open ASR LeaderboardBenchmarking Automatic Speech Recognition Tools for Iberian Languages (arXiv:2609.36920v1), Table 3
76Omnilingual ASR LLM 1B · bf16 Meta56±2local~6GBhow to run › Open ASR LeaderboardOmnilingual ASR: Open-Source Multilingual Speech Recognition for 1600+ Languages (arXiv:2511.09690v1)Open Universal Arabic ASR Leaderboard (Elm Research Center)Quranic ASR Leaderboard (Quran-Lab, held-out recitation set)
77Distil-Whisper distil-large-v3 · fp16 Hugging Face (Distil-Whisper)56±2local~2.3GBhow to run › distil-whisper/distil-large-v3.5 model cardOpen ASR Leaderboardhuggingface/distil-whisper README (Distil-Whisper headline WER table)
78Omnilingual ASR LLM 3B v2 · bf16 Meta56±3*local~11GBhow to run › Open ASR Leaderboard
79LiteWhisper large-v3 · fp16 Efficient-Speech (LiteASR)55±3local~3.7GBhow to run › LiteASR (arXiv:2502.20583v2) · ESB WER of Whisper and LiteASR-compressed encodersOpen ASR Leaderboard
80Fun-ASR-Nano-2512 · q8_0-gguf (transcribe.cpp) Alibaba (FunAudioLLM)55±3local~0.9GBhow to run › transcribe.cpp model catalog · LibriSpeech test-clean WER of validated full-precision portsmicrosoft/VibeVoice-ASR-BitNet model card (Accuracy table)
82LiteWhisper large-v3-acc · fp16 Efficient-Speech (LiteASR)55±3local~3.9GBhow to run › LiteASR (arXiv:2502.20583v2) · ESB WER of Whisper and LiteASR-compressed encodersOpen ASR Leaderboard
83Omnilingual ASR CTC 3B v2 · bf16 Meta55±3*local~7.9GBhow to run › Open ASR Leaderboard
84Omnilingual ASR CTC 7B · bf16 Meta55±2local (slow)~16GBhow to run › Open ASR LeaderboardOmnilingual ASR: Open-Source Multilingual Speech Recognition for 1600+ Languages (arXiv:2511.09690v1)Open Universal Arabic ASR Leaderboard (Elm Research Center)
85Parakeet CTC 1.1B · fp32 NVIDIA55±3local~5.6GBhow to run › Open ASR Leaderboardtranscribe.cpp model catalog · LibriSpeech test-clean WER of validated full-precision ports
86SeamlessM4T v2 Large · fp32 on CPU, fp16 on CUDA (seamless_communication m4t_predict) Meta55±4local~9.2GBhow to run › Canary-1B-v2 & Parakeet-TDT-0.6B-v3: Efficient and High-Performance Models for Multilingual ASR and AST (NVIDIA, arXiv:2509.14128v2)ai-sage/GigaAM-Multilingual model card (Model Performance table)Danish Speech-to-Text Leaderboard (RyeAI) · results datasetOpen Universal Arabic ASR Leaderboard (Elm Research Center)Turkish ASR Leaderboard (TurkASR-Bench, yagmurtuncer)Quranic ASR Leaderboard (Quran-Lab, held-out recitation set)+1
87Qwen3-ASR-0.6B · bf16 Alibaba (Qwen)55±1local~2.8GBhow to run › Qwen/Qwen3-ASR-1.7B model cardOpen ASR Leaderboardtranscribe.cpp model catalog · LibriSpeech test-clean WER of validated full-precision portsRW-Voice-EQ Bench (Hume AI)Persian ASR Triple Threat Leaderboard (Reza2kn)QCRI Arabic + English ASR leaderboard (Hugging Face Space QCRI/AURA-Arabic-English-ASR-leaderboard)+2
89Breeze ASR 25 · q8_0-gguf (transcribe.cpp) MediaTek Research54±4local~1.7GBhow to run › transcribe.cpp model catalog · LibriSpeech test-clean WER of validated full-precision ports
90Parakeet CTC 0.6B · fp32 NVIDIA54±3local~3.4GBhow to run › Open ASR Leaderboardtranscribe.cpp model catalog · LibriSpeech test-clean WER of validated full-precision ports
91Canary-180M-Flash · fp32 NVIDIA54±3local~1.4GBhow to run › transcribe.cpp model catalog · LibriSpeech test-clean WER of validated full-precision portsOpen ASR Leaderboard
92Omnilingual ASR CTC 3B · bf16 Meta54±2local~7.9GBhow to run › Open ASR LeaderboardOmnilingual ASR: Open-Source Multilingual Speech Recognition for 1600+ Languages (arXiv:2511.09690v1)Open Universal Arabic ASR Leaderboard (Elm Research Center)
93MOSS-Transcribe-Diarize 0.9B · q8_0-gguf (transcribe.cpp) OpenMOSS54±4local~1GBhow to run › transcribe.cpp model catalog · LibriSpeech test-clean WER of validated full-precision ports
95LiteWhisper large-v3-turbo-acc · fp16 Efficient-Speech (LiteASR)53±4*local~2.1GBhow to run › Open ASR LeaderboardLiteASR (arXiv:2502.20583v2) · ESB WER of Whisper and LiteASR-compressed encoders
96Voxtral Mini 4B Realtime (2602) · bf16 Mistral53±2local~11GBhow to run › Open ASR LeaderboardPipecat STT Benchmark · semantic WER results table (pipecat-ai/stt-benchmark README)Voxtral Realtimetranscribe.cpp model catalog · LibriSpeech test-clean WER of validated full-precision portsVibeVoice-ASR-Streaming Technical Report (arXiv:2609.02812v2), Table 3 single-speaker recognitionDutch ASR Leaderboard (tvosch) · committed result files+1
97Parakeet primeline (German) · q8_0-gguf (transcribe.cpp) primeLine53±4*local~0.7GBhow to run › transcribe.cpp model catalog · LibriSpeech test-clean WER of validated full-precision ports
98Omnilingual ASR CTC 7B v2 · bf16 Meta52±3*local (slow)~16GBhow to run › Open ASR Leaderboard
99Omnilingual ASR LLM 300M · bf16 Meta52±2local~4.4GBhow to run › Open ASR LeaderboardOmnilingual ASR: Open-Source Multilingual Speech Recognition for 1600+ Languages (arXiv:2511.09690v1)Open Universal Arabic ASR Leaderboard (Elm Research Center)Quranic ASR Leaderboard (Quran-Lab, held-out recitation set)
101LiteWhisper large-v3-fast · fp16 Efficient-Speech (LiteASR)52±3local~3.6GBhow to run › LiteASR (arXiv:2502.20583v2) · ESB WER of Whisper and LiteASR-compressed encodersOpen ASR Leaderboard
102VibeVoice-ASR-HF · bf16 Microsoft52±3local (slow)~21GBhow to run › microsoft/VibeVoice-ASR-BitNet model card (Accuracy table)Open ASR Leaderboard
103OWSM-CTC v4 1B · fp32 ESPnet (CMU WAVLab)52±4local~5.4GBhow to run › Open ASR LeaderboardOWSM v4 technical report (ESPnet / CMU WAVLab, arXiv:2506.00338v1): own OWSM-CTC scores
104Omnilingual ASR CTC 1B · bf16 Meta51±2local~2.9GBhow to run › Open ASR LeaderboardOmnilingual ASR: Open-Source Multilingual Speech Recognition for 1600+ Languages (arXiv:2511.09690v1)Open Universal Arabic ASR Leaderboard (Elm Research Center)
105Whisper large-v2 · fp32 OpenAI51±2local~7.9GBhow to run › transcribe.cpp model catalog · LibriSpeech test-clean WER of validated full-precision portsOpen ASR LeaderboardPESTE · Persian Speech to Text benchmark (fleurs-fa-ir-v1)German dialect ASR Benchmark (Betthupferl; kamaal-08/bavarian-asr-leaderboard)Open Universal Arabic ASR Leaderboard (Elm Research Center)Turkish ASR Leaderboard (TurkASR-Bench, yagmurtuncer)+1
106STT En Conformer-CTC Large · fp32 NVIDIA51±4*local~1.1GBhow to run › Open ASR Leaderboard
107Omnilingual ASR LLM 300M v2 · bf16 Meta51±3*local~4.4GBhow to run › Open ASR Leaderboard
108Moonshine Streaming Medium · fp32 Useful Sensors (Moonshine AI)50±3local~1.8GBhow to run › transcribe.cpp model catalog · LibriSpeech test-clean WER of validated full-precision portsOpen ASR LeaderboardPicovoice speech-to-text-benchmark · English batch-engine WER (README)
109VibeVoice-ASR-BitNet · I8_S VAE + I2_S LM (Q6_K embeddings) GGUF, VibeASR.cpp Microsoft50±4†local~1.7GBhow to run › microsoft/VibeVoice-ASR-BitNet model card (Accuracy table)
111Parakeet TDT-CTC 110M · fp32 NVIDIA50±3local~1.1GBhow to run › transcribe.cpp model catalog · LibriSpeech test-clean WER of validated full-precision portsOpen ASR Leaderboard
112ESPnet HuBERT Large + Conformer + Transformer LM · fp32 · CPU ESPnet50±4*†local~8GBhow to run › ESPnet LibriSpeech ASR recipe results (pinned September 2026)
113ESPnet WavLM Large + Conformer + Transformer LM · fp32 · CPU ESPnet50±4*†local~8GBhow to run › ESPnet LibriSpeech ASR recipe results (pinned September 2026)
114OWSM-CTC v3.1 1B · fp32 ESPnet (CMU WAVLab)50±4local~5.4GBhow to run › Open ASR LeaderboardOWSM v4 technical report (ESPnet / CMU WAVLab, arXiv:2506.00338v1): own OWSM-CTC scores
115Multitalker Parakeet Streaming 0.6B v1 · q8_0-gguf (transcribe.cpp) NVIDIA50±4*local~0.9GBhow to run › transcribe.cpp model catalog · LibriSpeech test-clean WER of validated full-precision ports
116Omnilingual ASR CTC 1B v2 · bf16 Meta50±3local~2.9GBhow to run › Open ASR LeaderboardGerman dialect ASR Benchmark (Betthupferl; kamaal-08/bavarian-asr-leaderboard)Benchmarking Automatic Speech Recognition Tools for Iberian Languages (arXiv:2609.36920v1), Table 3
117Whisper large (v1) · fp32 OpenAI49±2local~7.9GBhow to run › transcribe.cpp model catalog · LibriSpeech test-clean WER of validated full-precision portsOpen ASR LeaderboardOpen Universal Arabic ASR Leaderboard (Elm Research Center)AfriSpeech-200 (arXiv:2310.00274v1) · zero-shot WER of open ASR models on African-accented English
118OWSM-CTC v3.2 ft 1B · fp32 ESPnet (CMU WAVLab)49±4local~5.4GBhow to run › Open ASR LeaderboardOWSM v4 technical report (ESPnet / CMU WAVLab, arXiv:2506.00338v1): own OWSM-CTC scores
119Nemotron Speech Streaming EN 0.6B · q8_0-gguf (transcribe.cpp) NVIDIA49±4*local~0.7GBhow to run › transcribe.cpp model catalog · LibriSpeech test-clean WER of validated full-precision ports
121Whisper medium.en · fp32 OpenAI48±3local~4.2GBhow to run › transcribe.cpp model catalog · LibriSpeech test-clean WER of validated full-precision portsOpen ASR LeaderboardPicovoice speech-to-text-benchmark · English batch-engine WER (README)AfriSpeech-200 (arXiv:2310.00274v1) · zero-shot WER of open ASR models on African-accented EnglishSpeech Robust Bench (arXiv:2403.07937v3) · clean-speech WER of 25+ open ASR models
122Distil-Whisper distil-large-v2 · fp16 Hugging Face (Distil-Whisper)48±2local~2.3GBhow to run › distil-whisper/distil-large-v3.5 model cardOpen ASR Leaderboardhuggingface/distil-whisper README (Distil-Whisper headline WER table)
123Distil-Whisper distil-small.en · fp16 Hugging Face (Distil-Whisper)48±4*local~0.9GBhow to run › Open ASR Leaderboardhuggingface/distil-whisper README (Distil-Whisper headline WER table)
124CrisperWhisper · fp16 Nyra Health47±3local~4.4GBhow to run › AfriSpeech-MultiBench (arXiv:2511.14255v1) · African-accented English ASR benchmark suiteOpen ASR Leaderboard
125VibeVoice-ASR-Streaming-7B · bf16 (official VibeVoice repo) Microsoft47±5†local (slow)~17GBhow to run › VibeVoice-ASR-Streaming Technical Report (arXiv:2609.02812v2), Table 3 single-speaker recognition
126SpeechBrain Conformer (LargeScaleASR / Loquacious) · fp32 SpeechBrain47±5*local~2.8GBhow to run › Open ASR LeaderboardSpeechBrain Hugging Face ASR model cards: published LibriSpeech/VoxPopuli WER for its own checkpoints
127GigaAM Multilingual Large (CTC 600M) · fp32 (transformers remote code) Sber47±7*†local~2.3GBhow to run › ai-sage/GigaAM-Multilingual model card (Model Performance table)
128STT En Conformer-Transducer Large · fp32 NVIDIA46±4local~1.1GBhow to run › Open ASR LeaderboardNVIDIA NeMo docs: ASR scores tables (English Conformer / FastConformer)
129ESPnet E-Branchformer (no LM) · fp32 (espnet 202610.post2 + espnet_model_zoo 0.1.11) ESPnet46±4*†local~0.8GBhow to run › ESPnet LibriSpeech ASR recipe results (pinned September 2026)
130ESPnet Multiconvformer (no LM) · fp32 (espnet 202610.post2 + espnet_model_zoo 0.1.11) ESPnet46±4*†local~0.8GBhow to run › ESPnet LibriSpeech ASR recipe results (pinned September 2026)
131STT En FastConformer-Transducer Large · fp32 NVIDIA46±3local~1.1GBhow to run › Open ASR LeaderboardNVIDIA NeMo docs: ASR scores tables (English Conformer / FastConformer)
132Moonshine Streaming Small · fp32 Useful Sensors (Moonshine AI)46±3local~1.2GBhow to run › transcribe.cpp model catalog · LibriSpeech test-clean WER of validated full-precision portsOpen ASR LeaderboardPicovoice speech-to-text-benchmark · English batch-engine WER (README)
133SenseVoice Small · q8_0-gguf (transcribe.cpp) Alibaba (FunAudioLLM)46±3local~0.3GBhow to run › transcribe.cpp model catalog · LibriSpeech test-clean WER of validated full-precision portsmicrosoft/VibeVoice-ASR-BitNet model card (Accuracy table)
134STT En FastConformer-CTC Large · fp32 NVIDIA45±3local~1.1GBhow to run › Open ASR LeaderboardNVIDIA NeMo docs: ASR scores tables (English Conformer / FastConformer)
135ESPnet wav2vec 2.0 Large + Conformer + Transformer LM · fp32 · CPU ESPnet45±4*†local~8GBhow to run › ESPnet LibriSpeech ASR recipe results (pinned September 2026)
136ESPnet Conformer + S4 decoder (no LM) · fp32 · CPU ESPnet45±4*†local~8GBhow to run › ESPnet LibriSpeech ASR recipe results (pinned September 2026)
137Omnilingual ASR CTC 300M · bf16 Meta45±2local~1.3GBhow to run › Open ASR LeaderboardOmnilingual ASR: Open-Source Multilingual Speech Recognition for 1600+ Languages (arXiv:2511.09690v1)Open Universal Arabic ASR Leaderboard (Elm Research Center)
138Distil-Whisper distil-medium.en · fp16 Hugging Face (Distil-Whisper)44±3local~1.4GBhow to run › Open ASR Leaderboardhuggingface/distil-whisper README (Distil-Whisper headline WER table)
139Whisper small.en · fp32 OpenAI44±3local~1.7GBhow to run › transcribe.cpp model catalog · LibriSpeech test-clean WER of validated full-precision portsOpen ASR LeaderboardPicovoice speech-to-text-benchmark · English batch-engine WER (README)AfriSpeech-200 (arXiv:2310.00274v1) · zero-shot WER of open ASR models on African-accented EnglishSpeech Robust Bench (arXiv:2403.07937v3) · clean-speech WER of 25+ open ASR models
140ESPnet Branchformer 18-layer (70 epochs, no LM) · fp32 (espnet 202610.post2 + espnet_model_zoo 0.1.11) ESPnet44±4*†local~0.7GBhow to run › ESPnet LibriSpeech ASR recipe results (pinned September 2026)
141Nemotron 3.5 ASR Streaming 0.6B · fp32 (transformers >=5.13.0) NVIDIA44±3local~2.6GBhow to run › transcribe.cpp model catalog · LibriSpeech test-clean WER of validated full-precision portsVibeVoice-ASR-Streaming Technical Report (arXiv:2609.02812v2), Table 3 single-speaker recognitionPicovoice speech-to-text-benchmark · English batch-engine WER (README)Danish Speech-to-Text Leaderboard (RyeAI) · results datasetOpen Universal Arabic ASR Leaderboard (Elm Research Center)
142Whisper medium · fp32 OpenAI43±2local~4.2GBhow to run › transcribe.cpp model catalog · LibriSpeech test-clean WER of validated full-precision portsESPnet LibriSpeech ASR recipe results (pinned September 2026)Omnilingual ASR: Open-Source Multilingual Speech Recognition for 1600+ Languages (arXiv:2511.09690v1)LiteASR (arXiv:2502.20583v2) · ESB WER of Whisper and LiteASR-compressed encodersPESTE · Persian Speech to Text benchmark (fleurs-fa-ir-v1)German dialect ASR Benchmark (Betthupferl; kamaal-08/bavarian-asr-leaderboard)+4
143Omnilingual ASR CTC 300M v2 · bf16 Meta43±3*local~1.3GBhow to run › Open ASR Leaderboard
144STT En Conformer-Transducer Small · fp32 NVIDIA43±3local~0.6GBhow to run › Open ASR LeaderboardNVIDIA NeMo docs: ASR scores tables (English Conformer / FastConformer)
145ESPnet Conformer 2020 + RNN LM · fp32 (espnet 202610.post2 + espnet_model_zoo 0.1.11) ESPnet43±4*†local~1GBhow to run › ESPnet LibriSpeech ASR recipe results (pinned September 2026)
146Vosk en-us 0.22 · kaldi Alpha Cephei (Vosk)42±6local (slow)~16GBhow to run › Picovoice speech-to-text-benchmark · English batch-engine WER (README)Alpha Cephei Vosk models page (vendor-reported WER for its own Vosk models)
147GigaAM Multilingual (CTC 220M) · fp32 (transformers remote code) Sber42±7*†local~0.9GBhow to run › ai-sage/GigaAM-Multilingual model card (Model Performance table)
148ESPnet Transformer 18-layer + RNN LM · fp32 (espnet 202610.post2 + espnet_model_zoo 0.1.11) ESPnet42±4*†local~1GBhow to run › ESPnet LibriSpeech ASR recipe results (pinned September 2026)
149ABR asr-19m-v2-en · fp32 TorchScript (transformers remote code @ dab6545) Applied Brain Research41±5*local~0.1GBhow to run › Open ASR Leaderboard
151ESPnet Multi-blank Transducer 8421 · fp32 · CPU ESPnet40±4*†local~8GBhow to run › ESPnet LibriSpeech ASR recipe results (pinned September 2026)
152ESPnet Conformer RNN-T (no LM) · fp32 · CPU ESPnet40±4*†local~8GBhow to run › ESPnet LibriSpeech ASR recipe results (pinned September 2026)
153STT En Conformer-CTC Small · fp32 NVIDIA39±3local~0.6GBhow to run › Open ASR LeaderboardNVIDIA NeMo docs: ASR scores tables (English Conformer / FastConformer)
154Whisper small · fp32 OpenAI39±2local~1.7GBhow to run › transcribe.cpp model catalog · LibriSpeech test-clean WER of validated full-precision portsOmnilingual ASR: Open-Source Multilingual Speech Recognition for 1600+ Languages (arXiv:2511.09690v1)verliapp/stt-benchmark · independent STT accuracy benchmarkPESTE · Persian Speech to Text benchmark (fleurs-fa-ir-v1)Open ASR Leaderboard – Chilean Spanish (idsudd / Alonso Astroza)German dialect ASR Benchmark (Betthupferl; kamaal-08/bavarian-asr-leaderboard)+5
155Moonshine Base · fp32 Useful Sensors (Moonshine AI)38±3local~0.8GBhow to run › transcribe.cpp model catalog · LibriSpeech test-clean WER of validated full-precision portsOpen ASR Leaderboard
156Whisper base.en · fp32 OpenAI35±3local~0.9GBhow to run › transcribe.cpp model catalog · LibriSpeech test-clean WER of validated full-precision portsOpen ASR LeaderboardPicovoice speech-to-text-benchmark · English batch-engine WER (README)Speech Robust Bench (arXiv:2403.07937v3) · clean-speech WER of 25+ open ASR models
157SpeechBrain wav2vec2 CTC (LibriSpeech) · fp32 SpeechBrain35±4local~2GBhow to run › Open ASR LeaderboardSpeechBrain Hugging Face ASR model cards: published LibriSpeech/VoxPopuli WER for its own checkpoints
158Vosk small en-us 0.15 · kaldi Alpha Cephei (Vosk)31±6local~0.3GBhow to run › Picovoice speech-to-text-benchmark · English batch-engine WER (README)Alpha Cephei Vosk models page (vendor-reported WER for its own Vosk models)
159Moonshine Tiny · fp32 Useful Sensors (Moonshine AI)29±3local~0.6GBhow to run › transcribe.cpp model catalog · LibriSpeech test-clean WER of validated full-precision portsOpen ASR Leaderboard
160Moonshine Streaming Tiny · fp32 Useful Sensors (Moonshine AI)28±3local~0.7GBhow to run › transcribe.cpp model catalog · LibriSpeech test-clean WER of validated full-precision portsOpen ASR LeaderboardPicovoice speech-to-text-benchmark · English batch-engine WER (README)
161Whisper tiny.en · fp32 OpenAI27±3local~0.7GBhow to run › transcribe.cpp model catalog · LibriSpeech test-clean WER of validated full-precision portsOpen ASR LeaderboardPicovoice speech-to-text-benchmark · English batch-engine WER (README)Speech Robust Bench (arXiv:2403.07937v3) · clean-speech WER of 25+ open ASR models
162MMS-1B-all · fp32 Meta25±3local~5.1GBhow to run › Open ASR LeaderboardGerman dialect ASR Benchmark (Betthupferl; kamaal-08/bavarian-asr-leaderboard)Danish Speech-to-Text Leaderboard (RyeAI) · results datasetOpen Universal Arabic ASR Leaderboard (Elm Research Center)Turkish ASR Leaderboard (TurkASR-Bench, yagmurtuncer)Quranic ASR Leaderboard (Quran-Lab, held-out recitation set)
163Whisper base · fp32 OpenAI25±3local~0.9GBhow to run › transcribe.cpp model catalog · LibriSpeech test-clean WER of validated full-precision portsPESTE · Persian Speech to Text benchmark (fleurs-fa-ir-v1)German dialect ASR Benchmark (Betthupferl; kamaal-08/bavarian-asr-leaderboard)Danish Speech-to-Text Leaderboard (RyeAI) · results datasetTurkish ASR Leaderboard (TurkASR-Bench, yagmurtuncer)Speech Robust Bench (arXiv:2403.07937v3) · clean-speech WER of 25+ open ASR models
164wav2vec 2.0 Large 960h LV-60 self-training · fp32 Meta16±3local~2GBhow to run › Meta (facebook org) Hugging Face ASR model cards: published LibriSpeech WER for its own checkpointsOpen ASR LeaderboardAfriSpeech-200 (arXiv:2310.00274v1) · zero-shot WER of open ASR models on African-accented EnglishSpeech Robust Bench (arXiv:2403.07937v3) · clean-speech WER of 25+ open ASR models
165HuBERT X-Large (LibriSpeech 960h ft) · fp32 Meta16±4local~5.1GBhow to run › Open ASR LeaderboardMeta (facebook org) Hugging Face ASR model cards: published LibriSpeech WER for its own checkpointsAfriSpeech-200 (arXiv:2310.00274v1) · zero-shot WER of open ASR models on African-accented EnglishSpeech Robust Bench (arXiv:2403.07937v3) · clean-speech WER of 25+ open ASR models
166HuBERT Large (LibriSpeech 960h ft) · fp32 Meta15±4local~2GBhow to run › Open ASR LeaderboardMeta (facebook org) Hugging Face ASR model cards: published LibriSpeech WER for its own checkpointsSpeech Robust Bench (arXiv:2403.07937v3) · clean-speech WER of 25+ open ASR models
167wav2vec 2.0 Large Robust (LibriSpeech 960h ft) · fp32 Meta15±5local~2GBhow to run › Open ASR LeaderboardSpeech Robust Bench (arXiv:2403.07937v3) · clean-speech WER of 25+ open ASR modelsDysarthric ASR with clinical context on the Speech Accessibility Project (arXiv:2605.02782v2)
168Whisper tiny · fp32 OpenAI14±3local~0.7GBhow to run › transcribe.cpp model catalog · LibriSpeech test-clean WER of validated full-precision portsPESTE · Persian Speech to Text benchmark (fleurs-fa-ir-v1)German dialect ASR Benchmark (Betthupferl; kamaal-08/bavarian-asr-leaderboard)Danish Speech-to-Text Leaderboard (RyeAI) · results datasetTurkish ASR Leaderboard (TurkASR-Bench, yagmurtuncer)Speech Robust Bench (arXiv:2403.07937v3) · clean-speech WER of 25+ open ASR models
169wav2vec 2.0 Conformer Rel-Pos Large 960h · fp32 Meta14±4local~3.5GBhow to run › Meta (facebook org) Hugging Face ASR model cards: published LibriSpeech WER for its own checkpointsOpen ASR Leaderboard
170wav2vec 2.0 Conformer RoPE Large 960h · fp32 Meta13±4local~3.3GBhow to run › Meta (facebook org) Hugging Face ASR model cards: published LibriSpeech WER for its own checkpointsOpen ASR Leaderboard
171data2vec Audio Large 960h · fp32 Meta13±4local~2GBhow to run › Meta (facebook org) Hugging Face ASR model cards: published LibriSpeech WER for its own checkpointsOpen ASR Leaderboard
172wav2vec 2.0 Large 960h · fp32 Meta7±3local~2GBhow to run › Meta (facebook org) Hugging Face ASR model cards: published LibriSpeech WER for its own checkpointsOpen ASR LeaderboardAfriSpeech-200 (arXiv:2310.00274v1) · zero-shot WER of open ASR models on African-accented EnglishSpeech Robust Bench (arXiv:2403.07937v3) · clean-speech WER of 25+ open ASR modelsDysarthric ASR with clinical context on the Speech Accessibility Project (arXiv:2605.02782v2)
173data2vec Audio Base 960h · fp32 Meta4±4local~1GBhow to run › Meta (facebook org) Hugging Face ASR model cards: published LibriSpeech WER for its own checkpointsOpen ASR Leaderboard
174wav2vec 2.0 Base 960h · fp32 Meta2±3local~1GBhow to run › Meta (facebook org) Hugging Face ASR model cards: published LibriSpeech WER for its own checkpointsOpen ASR LeaderboardSpeech Robust Bench (arXiv:2403.07937v3) · clean-speech WER of 25+ open ASR models
175MMS-1B-FL102 · fp32 Meta-9±5local~5.1GBhow to run › Open ASR LeaderboardSpeech Robust Bench (arXiv:2403.07937v3) · clean-speech WER of 25+ open ASR models
176MedASR · q8_0-gguf (transcribe.cpp) Google-14±4*local~0.1GBhow to run › transcribe.cpp model catalog · LibriSpeech test-clean WER of validated full-precision ports

+ 50 more don't fit your machine or budget · show them

180 of 180 catalogue models listed here · sources and licences

* new or thinly tested: ranked on too few benchmark results · † only the maker’s own numbers so far · ~RAM is estimated; full context may need more memory