Voxtral Small 24B 2507
Jul 2025
Voxtral Small 24B 2507Compare inon
Voxtral Small 24B 2507 measured on 40 of 81 Speech to text benchmarks40
BenchmarkByScoreRank
No benchmark matches
Alternatives?
ModelScoreCostPriceWhat changes
MCheaperMultilingual+27%-83%$0.001ZBetterZoom Scribe V2 Pro+45%+11%$0.003Languages: not verified
OpenPhi 4 Multimodal Instruct+37%—$0.05Per audio minute: not verified
CheaperCanary Qwen 2.5B, NVIDIAOpensame-75%$0.001Languages: loses es, fr, de +4CheaperParakeet TDT 0.6B V3, TogetheraiOpensame-50%$0.002Memory (4-bit): not verifiedModelScoreCostPriceWhat changes
MMultilingual+27%-83%$0.001Canary Qwen 2.5B, NVIDIAOpensame-75%$0.001Languages: loses es, fr, de +4Parakeet TDT 0.6B V3, TogetheraiOpensame-50%$0.002Memory (4-bit): not verifiedVoxtral Mini 3B 2507Opensame-67%$0.001Memory (4-bit): not verified
Whisper Large v3 TurboOpen-18%-78%$0.001Memory (4-bit): not verifiedModelScoreCostPriceWhat changes
ZZoom Scribe V2 Pro+45%+11%$0.003Languages: not verifiedOSpeech v3.1+78%——Languages: not verified+1rResonant 1 Flash+71%——Languages: not verified+1ZZoom Scribe v1+66%——Languages: not verified+1Azure Speech 06 2026+62%——Languages: not verified+1ModelScoreCostPriceWhat changes
Phi 4 Multimodal InstructOpen+37%—$0.05Per audio minute: not verified
Granite Speech 3.3 8BOpen+31%——Languages: loses it, hi, nl+1Granite Speech 3.3 2BOpen+29%——Languages: loses it, hi, nl+1
Qwen 3 ASR 1.7B HfOpen+20%——Per audio minute: not verifiedBHiggs Audio v3 SttOpensame——Languages: loses es, fr, de +4+1ModelScoreCostPriceWhat changes
Voxtral Mini 4B Realtime 2602Open-23%——Per audio minute: not verifiedVoxtral Mini 3B 2507Opensame-67%$0.001Memory (4-bit): not verifiedVoxtral Mini Transcribe 3n/a—$0.004Languages: not verified+1Voxtral Small Transcribe 3n/a—$0.004Languages: not verified+1Voxtral Smalln/a—$0.004Languages: not verified+1Voxtral Mini Transcribe 2n/a—$0.003Voxtral Mini Transcribe Realtimen/a—$0.006Languages: not verified+1Voxtral Mini Transcribe Realtime 3n/a—$0.008Languages: not verified+1All Mistral models →Run it yourself
estimateFits on✗ 8 GB (phone, laptop)✓ 16 GB · compressed✓ 24 GB GPU · compressed✓ 48 GB GPU · 8-bit✓ 80 GB GPU · full quality
LicenceApache-2.0free for commercial use
Benchmarks
in Speech & VoiceOpen ASR · average WER6.9#37
Open ASR · RTFx101.5#44
Open ASR · AMI Cleaned WER13.2#48
Open ASR · URGENT 2024 WER12.4#23
Open ASR · AMI WER14.3#21
Open ASR · Earnings-22 WER10.2#20
Open ASR · VoxPopuli WER6.6#22
Open ASR · French RTFx89.2#20
Open ASR · Spanish RTFx101.5#20
Open ASR · Italian RTFx97.6#19
Open ASR · Portuguese RTFx122.8#19
Open ASR · Hindi Monsoon WER23.2#24
Open ASR · Dutch RTFx87.6#18
Open ASR · Dutch MLS WER5.6#13
FAQ
How much does Voxtral Small 24B 2507 cost?›
$0.003 per minute.
Is Voxtral Small 24B 2507 free to use?›
There is no free API; it costs $0.003 per minute. The weights are open, so you can run it yourself.
How good is Voxtral Small 24B 2507, and what is it best at?›
#37 of 61 in Speech to text.
Is Voxtral Small 24B 2507 better than Nemotron Speech Streaming En 0.6B?›
On Open ASR · average WER, Voxtral Small 24B 2507 scores 6.9 vs 6.9 — about the same.
How can I use Voxtral Small 24B 2507?›
Through the API of Mistral.
When was Voxtral Small 24B 2507 released? Is there a newer version?›
Released 2025-07-01. It is the current version in its line.
Is Voxtral Small 24B 2507 open source?›
Yes, the weights are open (Apache-2.0 licence).
Can I run Voxtral Small 24B 2507 locally?›
The weights are open and can be self-hosted; hardware needs are not confirmed yet.
What is Voxtral Small 24B 2507?›
Voxtral Small 24B 2507 is an open-weights speech & voice model from Mistral, released Jul 2025.
What are the best alternatives to Voxtral Small 24B 2507?›
Multilingual (cheaper), Zoom Scribe V2 Pro (better), Phi 4 Multimodal Instruct (open weights).