Model Menu
Models
Compare
Leaderboard
Benchmarks
⌘K
Top
Chat
Coding
Agents
Writing
Vision
Image
Video
Speech
Embeddings
Decision
Music
Filters
Best for Speech & Voice
All models
2292
Chat & Knowledge
558
Coding
381
Agents & Tool Use
541
Writing & Language
360
Vision & Documents
395
Image Generation & Editing
183
Video Generation
148
Speech & Voice
358
Embeddings & Retrieval
671
Decision models
118
Music & Sound
29
Text to speech
136
Speech to text
151
Realtime
55
Translation
8
Voice changers
4
AA Speech to Text
Open ASR
Open ASR · Armenian RTFx
All benchmarks
81
Speech & Voice benchmarks
81
×
Benchmark
By
Leader
Models
AA Speech to text
Recorded audio word error rate
AA Speech-to-Text · Non-streaming WER (version unstated): Measures transcription errors as a share of reference words in the non-streaming test
AA
Fun Realtime ASR Preview
1.7%
51
+3 versions & slices
Recorded audio word error rate · Telnyx
AA Speech-to-Text · Non-streaming WER (version unstated) · Telnyx · historical run: Measures transcription errors as a share of reference words in the non-streaming test
AA
D
Nova 3
4.8%
2
Recorded audio word error rate · fal.ai
AA Speech-to-Text · Non-streaming WER (version unstated) · fal.ai · historical run: Measures transcription errors as a share of reference words in the non-streaming test
AA
Whisper Large v3
4.1%
1
Recorded audio word error rate · Together AI
AA Speech-to-Text · Non-streaming WER (version unstated) · Together AI · historical run: Measures transcription errors as a share of reference words in the non-streaming test
AA
Whisper Large v3
4.5%
1
Open ASR
average WER
Open ASR · English Short Average WER: Share of words a transcription gets wrong on standard recordings
Open
O
Speech v3.1
3.9%
61
+76 versions & slices
Average WER · GPU ASR / greedy CTC
Open ASR · English Short Average WER · Open ASR Leaderboard · historical run: Share of words a transcription gets wrong on standard recordings
Open
Parakeet TDT 0.6B V3, Togetherai
6.4%
3
AMI Cleaned WER
Open ASR · English Short AMI Cleaned WER: Share of words a transcription gets wrong on standard recordings
Open
Z
Zoom Scribe V2 Pro
6.1%
61
Earnings-22 Cleaned AA Chunked WER
Open ASR · English Short Earnings-22 Cleaned AA Chunked WER: Share of words a transcription gets wrong on standard recordings
Open
Z
Zoom Scribe V2 Pro
4.1%
61
GigaSpeech Cleaned WER
Open ASR · English Short GigaSpeech Cleaned WER: Share of words a transcription gets wrong on standard recordings
Open
H
Hojo ASR v1
6.1%
61
LibriSpeech Clean WER
Open ASR · English Short LibriSpeech Clean WER: Share of words a transcription gets wrong on standard recordings
Open
M
Multilingual
0.9%
61
LibriSpeech Other WER
Open ASR · English Short LibriSpeech Other WER: Share of words a transcription gets wrong on standard recordings
Open
M
Multilingual
1.8%
61
SPGISpeech WER
Open ASR · English Short Spgispeech WER: Share of words a transcription gets wrong on standard recordings
Open
Z
Zoom Scribe V2 Pro
1.4%
61
Voice Arena Monsoon WER
Open ASR · English Short Voice Arena Monsoon WER: Share of words a transcription gets wrong on standard recordings
Open
Z
Zoom Scribe V2 Pro
3%
61
VoxPopuli AA Cleaned WER
Open ASR · English Short VoxPopuli AA Cleaned WER: Share of words a transcription gets wrong on standard recordings
Open
Scribe v2
1.4%
61
URGENT 2024 Clean WER
Open ASR · English Short Urgent2024 Clean WER: Share of words a transcription gets wrong on standard recordings
Open
O
Speech v3.1
3.9%
53
URGENT 2024 WER
Open ASR · English Short Urgent2024 WER: Share of words a transcription gets wrong on standard recordings
Open
Z
Zoom Scribe V2 Pro
8%
53
RTFx
Open ASR · English Short RTFx: This measures how much audio a model transcribes per unit of processing time
Open
Granite Speech 5.0 470M Turboctc
21,134
45
Earnings-22 WER
Open ASR · English Short Earnings-22 WER: Share of words a transcription gets wrong on standard recordings
Open
Azure Speech 06 2026
7.5%
42
German FLEURS WER
Open ASR · German FLEURS WER: Share of words a transcription gets wrong on standard recordings
Open
Azure Speech 07 2026
1.8%
40
German MCV WER
Open ASR · German MCV WER: Share of words a transcription gets wrong on standard recordings
Open
Azure Speech 06 2026
1.9%
40
French FLEURS WER
Open ASR · French FLEURS WER: Share of words a transcription gets wrong on standard recordings
Open
Azure Speech 07 2026
2.5%
39
French MCV WER
Open ASR · French MCV WER: Share of words a transcription gets wrong on standard recordings
Open
Azure Speech 06 2026
4.3%
39
French MLS WER
Open ASR · French MLS WER: Share of words a transcription gets wrong on standard recordings
Open
A
Universal 3 Pro
2.3%
39
Portuguese FLEURS WER
Open ASR · Portuguese FLEURS WER: Share of words a transcription gets wrong on standard recordings
Open
Azure Speech 07 2026
2.1%
39
Portuguese MLS WER
Open ASR · Portuguese MLS WER: Share of words a transcription gets wrong on standard recordings
Open
Scribe v2
3.2%
39
Spanish FLEURS WER
Open ASR · Spanish FLEURS WER: Share of words a transcription gets wrong on standard recordings
Open
Azure Speech 06 2026
1.8%
39
Spanish MCV WER
Open ASR · Spanish MCV WER: Share of words a transcription gets wrong on standard recordings
Open
S
Speechmatics Enhanced
2.1%
39
Spanish MLS WER
Open ASR · Spanish MLS WER: Share of words a transcription gets wrong on standard recordings
Open
C
Cohere Transcribe 03 2026
2.4%
39
Italian FLEURS WER
Open ASR · Italian FLEURS WER: Share of words a transcription gets wrong on standard recordings
Open
Azure Speech 07 2026
0.9%
38
Italian MCV WER
Open ASR · Italian MCV WER: Share of words a transcription gets wrong on standard recordings
Open
Azure Speech 06 2026
1.8%
38
Italian MLS WER
Open ASR · Italian MLS WER: Share of words a transcription gets wrong on standard recordings
Open
M
Multilingual
4.1%
38
Dutch FLEURS WER
Open ASR · Dutch FLEURS WER: Share of words a transcription gets wrong on standard recordings
Open
Scribe v2
2.5%
35
Dutch MCV WER
Open ASR · Dutch MCV WER: Share of words a transcription gets wrong on standard recordings
Open
Azure Speech 06 2026
1.5%
35
Dutch MLS WER
Open ASR · Dutch MLS WER: Share of words a transcription gets wrong on standard recordings
Open
M
Multilingual
3.5%
35
AMI WER
Open ASR · English Short AMI WER: Share of words a transcription gets wrong on standard recordings
Open
Granite Speech 4.1 2B Nar
7.7%
32
GigaSpeech WER
Open ASR · English Short GigaSpeech WER: Share of words a transcription gets wrong on standard recordings
Open
E
ARK ASR 3B
7%
32
VoxPopuli WER
Open ASR · English Short VoxPopuli WER: Share of words a transcription gets wrong on standard recordings
Open
Azure Speech 06 2026
5.1%
32
Hindi Monsoon WER
Open ASR · Hindi Monsoon WER: Share of words a transcription gets wrong on standard recordings
Open
Muse Voice Transcribe
6.9%
32
French RTFx
Open ASR · French RTFx: This measures how much audio a model transcribes per unit of processing time
Open
Parakeet TDT 0.6B V3, Togetherai
3,618
26
German RTFx
Open ASR · German RTFx: This measures how much audio a model transcribes per unit of processing time
Open
Parakeet TDT 0.6B V3, Togetherai
3,363
26
Portuguese RTFx
Open ASR · Portuguese RTFx: This measures how much audio a model transcribes per unit of processing time
Open
Parakeet TDT 0.6B V3, Togetherai
3,257
26
Spanish RTFx
Open ASR · Spanish RTFx: This measures how much audio a model transcribes per unit of processing time
Open
Parakeet TDT 0.6B V3, Togetherai
3,786
26
Italian RTFx
Open ASR · Italian RTFx: This measures how much audio a model transcribes per unit of processing time
Open
Parakeet TDT 0.6B V3, Togetherai
3,427
25
Dutch RTFx
Open ASR · Dutch RTFx: This measures how much audio a model transcribes per unit of processing time
Open
Parakeet TDT 0.6B V3, Togetherai
2,783
23
Hindi RTFx
Open ASR · Hindi RTFx: This measures how much audio a model transcribes per unit of processing time
Open
OmniASR CTC 300M v2
971
21
Armenian FLEURS WER
Open ASR · Armenian FLEURS WER: Share of words a transcription gets wrong on standard recordings
Open
Scribe v2
3.2%
15
Armenian MCV WER
Open ASR · Armenian MCV WER: Share of words a transcription gets wrong on standard recordings
Open
Stt Hy Fastconformer Hybrid Large Pc
4.9%
15
Armenian RTFx
Open ASR · Armenian RTFx: This measures how much audio a model transcribes per unit of processing time
Open
Stt Hy Fastconformer Hybrid Large Pc
2,879
10
RTFx · GPU ASR / greedy CTC
Open ASR · English Short RTFx · Open ASR Leaderboard · historical run: This measures how much audio a model transcribes per unit of processing time
Open
s
Zipformer-cr-ctc-transducer-XL-290M (fast-gpu-asr, transducer_modified_beam_search)
20,886
3
AMI Cleaned WER · GPU ASR / greedy CTC
Open ASR · English Short AMI Cleaned WER · Open ASR Leaderboard · historical run: Share of words a transcription gets wrong on standard recordings
Open
Parakeet TDT 0.6B V2, NVIDIA
9%
3
Earnings-22 Cleaned AA Chunked WER · GPU ASR / greedy CTC
Open ASR · English Short Earnings-22 Cleaned AA Chunked WER · Open ASR Leaderboard · historical run: Share of words a transcription gets wrong on standard recordings
Open
Parakeet TDT 0.6B V3, Togetherai
5.8%
3
GigaSpeech Cleaned WER · GPU ASR / greedy CTC
Open ASR · English Short GigaSpeech Cleaned WER · Open ASR Leaderboard · historical run: Share of words a transcription gets wrong on standard recordings
Open
Parakeet TDT 0.6B V3, Togetherai
7.8%
3
LibriSpeech Clean WER · GPU ASR / greedy CTC
Open ASR · English Short LibriSpeech Clean WER · Open ASR Leaderboard · historical run: Share of words a transcription gets wrong on standard recordings
Open
Parakeet TDT 0.6B V2, NVIDIA
1.3%
3
LibriSpeech Other WER · GPU ASR / greedy CTC
Open ASR · English Short LibriSpeech Other WER · Open ASR Leaderboard · historical run: Share of words a transcription gets wrong on standard recordings
Open
Parakeet TDT 0.6B V2, NVIDIA
2.7%
3
SPGISpeech WER · GPU ASR / greedy CTC
Open ASR · English Short Spgispeech WER · Open ASR Leaderboard · historical run: Share of words a transcription gets wrong on standard recordings
Open
Parakeet TDT 0.6B V2, NVIDIA
1.9%
3
URGENT 2024 WER · GPU ASR / greedy CTC
Open ASR · English Short Urgent2024 WER · Open ASR Leaderboard · historical run: Share of words a transcription gets wrong on standard recordings
Open
Parakeet TDT 0.6B V3, Togetherai
11.5%
3
Voice Arena Monsoon WER · GPU ASR / greedy CTC
Open ASR · English Short Voice Arena Monsoon WER · Open ASR Leaderboard · historical run: Share of words a transcription gets wrong on standard recordings
Open
Parakeet TDT 0.6B V2, NVIDIA
3.7%
3
VoxPopuli AA Cleaned WER · GPU ASR / greedy CTC
Open ASR · English Short VoxPopuli AA Cleaned WER · Open ASR Leaderboard · historical run: Share of words a transcription gets wrong on standard recordings
Open
Parakeet TDT 0.6B V3, Togetherai
3%
3
URGENT 2024 Clean WER · GPU ASR / greedy CTC
Open ASR · English Short Urgent2024 Clean WER · Open ASR Leaderboard · historical run: Share of words a transcription gets wrong on standard recordings
Open
Parakeet TDT 0.6B V3, Togetherai
5.3%
3
French RTFx · GPU ASR
Open ASR · French RTFx · Open ASR Leaderboard · historical run: This measures how much audio a model transcribes per unit of processing time
Open
Parakeet TDT 0.6B V3, Togetherai
10,882
1
French FLEURS WER · GPU ASR
Open ASR · French FLEURS WER · Open ASR Leaderboard · historical run: Share of words a transcription gets wrong on standard recordings
Open
Parakeet TDT 0.6B V3, Togetherai
4.7%
1
French MCV WER · GPU ASR
Open ASR · French MCV WER · Open ASR Leaderboard · historical run: Share of words a transcription gets wrong on standard recordings
Open
Parakeet TDT 0.6B V3, Togetherai
6.2%
1
French MLS WER · GPU ASR
Open ASR · French MLS WER · Open ASR Leaderboard · historical run: Share of words a transcription gets wrong on standard recordings
Open
Parakeet TDT 0.6B V3, Togetherai
5%
1
German RTFx · GPU ASR
Open ASR · German RTFx · Open ASR Leaderboard · historical run: This measures how much audio a model transcribes per unit of processing time
Open
Parakeet TDT 0.6B V3, Togetherai
10,862
1
German FLEURS WER · GPU ASR
Open ASR · German FLEURS WER · Open ASR Leaderboard · historical run: Share of words a transcription gets wrong on standard recordings
Open
Parakeet TDT 0.6B V3, Togetherai
4.1%
1
German MCV WER · GPU ASR
Open ASR · German MCV WER · Open ASR Leaderboard · historical run: Share of words a transcription gets wrong on standard recordings
Open
Parakeet TDT 0.6B V3, Togetherai
3.9%
1
Spanish RTFx · GPU ASR
Open ASR · Spanish RTFx · Open ASR Leaderboard · historical run: This measures how much audio a model transcribes per unit of processing time
Open
Parakeet TDT 0.6B V3, Togetherai
11,582
1
Spanish FLEURS WER · GPU ASR
Open ASR · Spanish FLEURS WER · Open ASR Leaderboard · historical run: Share of words a transcription gets wrong on standard recordings
Open
Parakeet TDT 0.6B V3, Togetherai
3.1%
1
Spanish MCV WER · GPU ASR
Open ASR · Spanish MCV WER · Open ASR Leaderboard · historical run: Share of words a transcription gets wrong on standard recordings
Open
Parakeet TDT 0.6B V3, Togetherai
3.4%
1
Spanish MLS WER · GPU ASR
Open ASR · Spanish MLS WER · Open ASR Leaderboard · historical run: Share of words a transcription gets wrong on standard recordings
Open
Parakeet TDT 0.6B V3, Togetherai
4.3%
1
Italian RTFx · GPU ASR
Open ASR · Italian RTFx · Open ASR Leaderboard · historical run: This measures how much audio a model transcribes per unit of processing time
Open
Parakeet TDT 0.6B V3, Togetherai
11,165
1
Italian FLEURS WER · GPU ASR
Open ASR · Italian FLEURS WER · Open ASR Leaderboard · historical run: Share of words a transcription gets wrong on standard recordings
Open
Parakeet TDT 0.6B V3, Togetherai
2.4%
1
Italian MCV WER · GPU ASR
Open ASR · Italian MCV WER · Open ASR Leaderboard · historical run: Share of words a transcription gets wrong on standard recordings
Open
Parakeet TDT 0.6B V3, Togetherai
3.3%
1
Italian MLS WER · GPU ASR
Open ASR · Italian MLS WER · Open ASR Leaderboard · historical run: Share of words a transcription gets wrong on standard recordings
Open
Parakeet TDT 0.6B V3, Togetherai
8.1%
1
Portuguese RTFx · GPU ASR
Open ASR · Portuguese RTFx · Open ASR Leaderboard · historical run: This measures how much audio a model transcribes per unit of processing time
Open
Parakeet TDT 0.6B V3, Togetherai
10,147
1
Portuguese FLEURS WER · GPU ASR
Open ASR · Portuguese FLEURS WER · Open ASR Leaderboard · historical run: Share of words a transcription gets wrong on standard recordings
Open
Parakeet TDT 0.6B V3, Togetherai
4.4%
1
Portuguese MLS WER · GPU ASR
Open ASR · Portuguese MLS WER · Open ASR Leaderboard · historical run: Share of words a transcription gets wrong on standard recordings
Open
Parakeet TDT 0.6B V3, Togetherai
8%
1
Dutch RTFx · GPU ASR
Open ASR · Dutch RTFx · Open ASR Leaderboard · historical run: This measures how much audio a model transcribes per unit of processing time
Open
Parakeet TDT 0.6B V3, Togetherai
10,391
1
Dutch FLEURS WER · GPU ASR
Open ASR · Dutch FLEURS WER · Open ASR Leaderboard · historical run: Share of words a transcription gets wrong on standard recordings
Open
Parakeet TDT 0.6B V3, Togetherai
6.3%
1
Dutch MCV WER · GPU ASR
Open ASR · Dutch MCV WER · Open ASR Leaderboard · historical run: Share of words a transcription gets wrong on standard recordings
Open
Parakeet TDT 0.6B V3, Togetherai
5.3%
1
Dutch MLS WER · GPU ASR
Open ASR · Dutch MLS WER · Open ASR Leaderboard · historical run: Share of words a transcription gets wrong on standard recordings
Open
Parakeet TDT 0.6B V3, Togetherai
10.8%
1
No benchmark matches
Model
Open ASR · Armenian RTFx
Languages
Price
Released
Stt Hy Fastconformer Hybrid Large Pc
NVIDIA
Open
2,879.4
hy
Nov 2024
OmniASR CTC 300M v2
Meta
1,084.2
—
OmniASR CTC 1B v2
Meta
763.9
—
OmniASR CTC 3B v2
Meta
527.6
—
OmniASR CTC 7B v2
Meta
310.9
—
Whisper Large v3
OpenAI
Open
73.4
en, es, fr +86
$0.004
/ minute
Sep 2022
OmniASR LLM 300M v2
Meta
47.7
—
OmniASR LLM 1B v2
Meta
46.8
—
OmniASR LLM 3B v2
Meta
45.2
—
OmniASR LLM 7B v2
Meta
42.6
—
Show 141 not measured on Open ASR · Armenian RTFx
