Audio modelPublic preview
Voxtral Mini TTS 3
Mistral AI
- Released
- October 20, 2026
- Data date
- October 24, 2026
- Languages
Voxtral Mini TTS 3 generates speech from text and supports reference-based voice cloning. Mistral lists streaming, time to first audio of about 90 milliseconds, and five output formats. Published language counts differ by source.
Specifications and access
| Specification | Value and source |
|---|---|
| API model ID | voxtral-mini-tts-3-0Source |
| Release | Public preview since October 20, 2026Source |
| Access | Mistral API, public previewSource |
| Input | Text; voice cloning can use a speech referenceSource |
| Output | Generated speech; five output formatsSource |
| Languages | Model card: 9 languages. Product announcement: more than 35 languages and over 50 dialects.Source |
| Time to first audio | Approximately 90 millisecondsSource |
| Voice controls | 34 emotion tags and two pace tagsSource |
| Release | 2026-10-20Source |
| Languages in the model card | |
| Languages in the announcement | 35+ languages and 50+ dialectsSource |
| Time to first audio in milliseconds | 90Source |
| Output formats | |
| Emotion tags | 34Source |
| Pace tags | |
| Price per million characters in USD | 13Source |
| Cached price per million characters in USD | 1.3Source |
| VoiceArena language coverage | Methodology says this exact model is listed only for US English and Hindi.Source |
| Language support in the current documentation | Current model-specific table reports 35 languages and 50+ dialects.Source |
| Preset voice languages | Arabic (ar); Catalan (ca); Chinese (zh); Dutch (nl); English UK (en_GB); English US (en_us); French (fr); German (de); Hindi (hi); Hungarian (hu); Italian (it); Japanese (ja); Korean (ko); Norwegian (no); Polish (pl); Portuguese (pt); Romanian (ro); Russian (ru); Spanish (es); Turkish (tr); Ukrainian (uk); Vietnamese (vi). This lists preset-voice coverage; overall language support is broader.Source |
| Documented model mode | Official TTS overview labels voxtral-mini-tts-3-0 Offline.Source |
| Streaming audio output behavior | The stream guide says all text is sent at once and generated audio is returned in chunks.Source |
| Jointly documented model IDs | The shared names array lists voxtral-mini-tts-3-0, voxtral-mini-tts-3, voxtral-mini-tts-latest, voxtral-mini-tts-realtime-3-0, voxtral-mini-tts-realtime-3, voxtral-mini-tts-realtime-latest. (A shared names array is not an explicit identity, alias, checkpoint, release, price or benchmark-configuration bridge; no scores or prices are transferred between offline and realtime TTS.)Source |
Benchmark sources at a glance
Each evaluator measures different capabilities under its own conditions. Test series, composite indices and their components stay separate. The detailed results below retain their scales, configurations and source dates.
No active result from the reviewed external evaluators can currently be attributed to this model. Other source evidence remains available below.
Coverage is not a quality ranking. Overlapping tests and multiple settings do not count as independent experiments. Vendor claims, historical and disputed results are excluded from this overview and remain labelled in the detailed evidence.
Page 1 of 2
Measurements without matching peer values
These measurements have no matching peer values under the same test conditions. Their original values and sources remain available here.
Voice Code Bench faithful clips (Show measurement, test conditions, and source)
- Source value
- 79.6667
- Score
- 79.6667
- Metric
- faithful clip rate
- Unit
- %
- Category
- speech generation
- Direction
- Higher is better
- Sample
- 300
- Source type
- vendor-reported
- Evaluator
- Mistral Voxtral 3 announcement and benchmark tables
- Status
- active
- Retrieved at
- 2026-10-24
- Methodology
- Vendor-reported result. 300 English prompts, two voices (female and male); BeSimple AI evaluation; Gemini 3.8 Flash judge; chart x-axis lists USD 13 per million characters.
Pronunciation control (Show measurement, test conditions, and source)
- Source value
- 86.83
- Score
- 86.83
- Metric
- correct pronunciation of target word
- Unit
- %
- Category
- speech generation
- Direction
- Higher is better
- Sample
- 300
- Source type
- vendor-reported
- Evaluator
- Mistral Voxtral 3 announcement and benchmark tables
- Status
- active
- Retrieved at
- 2026-10-24
- Methodology
- Vendor-reported result. 300 English sentences, 150 easy and 150 hard; two voices; IPA target for a rare or made-up word; Gemini 3.8 Flash judge.
VoiceArena TTS source-reported win_rate: US English, All (Show measurement, test conditions, and source)
- Source value
- 51.84
- Score
- 51.84
- Metric
- source-reported win_rate
- Unit
- %
- Benchmark version
- Current public leaderboard captured 2026-10-24
- Category
- tts-preference
- Direction
- Higher is better
- Source type
- official-leaderboard
- Evaluator
- VoiceArena US English TTS leaderboard
- Status
- active
- Retrieved at
- 2026-10-24
- Methodology
- Official benchmark leaderboard. VoiceArena US English-only board, category "All"; source-reported win_rate 51.84%; rank 10, bootstrap rank range 7-15. The source separately lists 733 pairwise battles. Higher win_rate means a higher reported win share. Compare only within the same language and category.
- Context
- The source labels this field win_rate but does not publish its formula or denominator. The separately listed pairwise-battle count is not identified as the win_rate denominator. No Elo confidence interval applies to this metric.
VoiceArena TTS source-reported win_rate: US English, Content creation & Edu (Show measurement, test conditions, and source)
- Source value
- 50
- Score
- 50
- Metric
- source-reported win_rate
- Unit
- %
- Benchmark version
- Current public leaderboard captured 2026-10-24
- Category
- tts-preference
- Direction
- Higher is better
- Source type
- official-leaderboard
- Evaluator
- VoiceArena US English TTS leaderboard
- Status
- active
- Retrieved at
- 2026-10-24
- Methodology
- Official benchmark leaderboard. VoiceArena US English-only board, category "Content creation & Edu"; source-reported win_rate 50.0%; rank 14, bootstrap rank range 4-21. The source separately lists 145 pairwise battles. Higher win_rate means a higher reported win share. Compare only within the same language and category.
- Context
- The source labels this field win_rate but does not publish its formula or denominator. The separately listed pairwise-battle count is not identified as the win_rate denominator. No Elo confidence interval applies to this metric.
VoiceArena TTS source-reported win_rate: US English, Customer support (Show measurement, test conditions, and source)
- Source value
- 49.63
- Score
- 49.63
- Metric
- source-reported win_rate
- Unit
- %
- Benchmark version
- Current public leaderboard captured 2026-10-24
- Category
- tts-preference
- Direction
- Higher is better
- Source type
- official-leaderboard
- Evaluator
- VoiceArena US English TTS leaderboard
- Status
- active
- Retrieved at
- 2026-10-24
- Methodology
- Official benchmark leaderboard. VoiceArena US English-only board, category "Customer support"; source-reported win_rate 49.63%; rank 12, bootstrap rank range 6-18. The source separately lists 267 pairwise battles. Higher win_rate means a higher reported win share. Compare only within the same language and category.
- Context
- The source labels this field win_rate but does not publish its formula or denominator. The separately listed pairwise-battle count is not identified as the win_rate denominator. No Elo confidence interval applies to this metric.
VoiceArena TTS source-reported win_rate: US English, Gen-conv (Show measurement, test conditions, and source)
- Source value
- 54.87
- Score
- 54.87
- Metric
- source-reported win_rate
- Unit
- %
- Benchmark version
- Current public leaderboard captured 2026-10-24
- Category
- tts-preference
- Direction
- Higher is better
- Source type
- official-leaderboard
- Evaluator
- VoiceArena US English TTS leaderboard
- Status
- active
- Retrieved at
- 2026-10-24
- Methodology
- Official benchmark leaderboard. VoiceArena US English-only board, category "Gen-conv"; source-reported win_rate 54.87%; rank 6, bootstrap rank range 2-17. The source separately lists 154 pairwise battles. Higher win_rate means a higher reported win share. Compare only within the same language and category.
- Context
- The source labels this field win_rate but does not publish its formula or denominator. The separately listed pairwise-battle count is not identified as the win_rate denominator. No Elo confidence interval applies to this metric.
VoiceArena TTS source-reported win_rate: US English, Media & Entertainment (Show measurement, test conditions, and source)
- Source value
- 54.19
- Score
- 54.19
- Metric
- source-reported win_rate
- Unit
- %
- Benchmark version
- Current public leaderboard captured 2026-10-24
- Category
- tts-preference
- Direction
- Higher is better
- Source type
- official-leaderboard
- Evaluator
- VoiceArena US English TTS leaderboard
- Status
- active
- Retrieved at
- 2026-10-24
- Methodology
- Official benchmark leaderboard. VoiceArena US English-only board, category "Media & Entertainment"; source-reported win_rate 54.19%; rank 10, bootstrap rank range 1-17. The source separately lists 167 pairwise battles. Higher win_rate means a higher reported win share. Compare only within the same language and category.
- Context
- The source labels this field win_rate but does not publish its formula or denominator. The separately listed pairwise-battle count is not identified as the win_rate denominator. No Elo confidence interval applies to this metric.
VoiceArena TTS source-reported win_rate: Hindi, All (Show measurement, test conditions, and source)
- Source value
- 49.11
- Score
- 49.11
- Metric
- source-reported win_rate
- Unit
- %
- Benchmark version
- Current public leaderboard captured 2026-10-24
- Category
- tts-preference
- Direction
- Higher is better
- Source type
- official-leaderboard
- Evaluator
- VoiceArena Hindi TTS leaderboard
- Status
- active
- Retrieved at
- 2026-10-24
- Methodology
- Official benchmark leaderboard. VoiceArena Hindi-only board, category "All"; source-reported win_rate 49.11%; rank 6, bootstrap rank range 5-11. The source separately lists 616 pairwise battles. Higher win_rate means a higher reported win share. Compare only within the same language and category.
- Context
- The source labels this field win_rate but does not publish its formula or denominator. The separately listed pairwise-battle count is not identified as the win_rate denominator. No Elo confidence interval applies to this metric.
VoiceArena TTS source-reported win_rate: Hindi, Content creation & Edu (Show measurement, test conditions, and source)
- Source value
- 56.5
- Score
- 56.5
- Metric
- source-reported win_rate
- Unit
- %
- Benchmark version
- Current public leaderboard captured 2026-10-24
- Category
- tts-preference
- Direction
- Higher is better
- Source type
- official-leaderboard
- Evaluator
- VoiceArena Hindi TTS leaderboard
- Status
- active
- Retrieved at
- 2026-10-24
- Methodology
- Official benchmark leaderboard. VoiceArena Hindi-only board, category "Content creation & Edu"; source-reported win_rate 56.5%; rank 3, bootstrap rank range 1-10. The source separately lists 123 pairwise battles. Higher win_rate means a higher reported win share. Compare only within the same language and category.
- Context
- The source labels this field win_rate but does not publish its formula or denominator. The separately listed pairwise-battle count is not identified as the win_rate denominator. No Elo confidence interval applies to this metric.
VoiceArena TTS source-reported win_rate: Hindi, Customer support (Show measurement, test conditions, and source)
- Source value
- 47.77
- Score
- 47.77
- Metric
- source-reported win_rate
- Unit
- %
- Benchmark version
- Current public leaderboard captured 2026-10-24
- Category
- tts-preference
- Direction
- Higher is better
- Source type
- official-leaderboard
- Evaluator
- VoiceArena Hindi TTS leaderboard
- Status
- active
- Retrieved at
- 2026-10-24
- Methodology
- Official benchmark leaderboard. VoiceArena Hindi-only board, category "Customer support"; source-reported win_rate 47.77%; rank 8, bootstrap rank range 4-12. The source separately lists 224 pairwise battles. Higher win_rate means a higher reported win share. Compare only within the same language and category.
- Context
- The source labels this field win_rate but does not publish its formula or denominator. The separately listed pairwise-battle count is not identified as the win_rate denominator. No Elo confidence interval applies to this metric.
VoiceArena TTS source-reported win_rate: Hindi, Gen-conv (Show measurement, test conditions, and source)
- Source value
- 49.6
- Score
- 49.6
- Metric
- source-reported win_rate
- Unit
- %
- Benchmark version
- Current public leaderboard captured 2026-10-24
- Category
- tts-preference
- Direction
- Higher is better
- Source type
- official-leaderboard
- Evaluator
- VoiceArena Hindi TTS leaderboard
- Status
- active
- Retrieved at
- 2026-10-24
- Methodology
- Official benchmark leaderboard. VoiceArena Hindi-only board, category "Gen-conv"; source-reported win_rate 49.6%; rank 6, bootstrap rank range 1-13. The source separately lists 126 pairwise battles. Higher win_rate means a higher reported win share. Compare only within the same language and category.
- Context
- The source labels this field win_rate but does not publish its formula or denominator. The separately listed pairwise-battle count is not identified as the win_rate denominator. No Elo confidence interval applies to this metric.
VoiceArena TTS source-reported win_rate: Hindi, Media & Entertainment (Show measurement, test conditions, and source)
- Source value
- 44.41
- Score
- 44.41
- Metric
- source-reported win_rate
- Unit
- %
- Benchmark version
- Current public leaderboard captured 2026-10-24
- Category
- tts-preference
- Direction
- Higher is better
- Source type
- official-leaderboard
- Evaluator
- VoiceArena Hindi TTS leaderboard
- Status
- active
- Retrieved at
- 2026-10-24
- Methodology
- Official benchmark leaderboard. VoiceArena Hindi-only board, category "Media & Entertainment"; source-reported win_rate 44.41%; rank 10, bootstrap rank range 5-16. The source separately lists 143 pairwise battles. Higher win_rate means a higher reported win share. Compare only within the same language and category.
- Context
- The source labels this field win_rate but does not publish its formula or denominator. The separately listed pairwise-battle count is not identified as the win_rate denominator. No Elo confidence interval applies to this metric.
Page 1 of 2
Published prices
- Price
- USD 13 per 1 million characters; cached rate USD 1.30 per 1 million charactersSource
Sources and data date
| Type | Evidence and data date |
|---|---|
| Research status | Research date October 24, 2026. 93 source URLs checked. This documents the inspected sources, not an exhaustive inventory of every publication. |
| Unresolved | Supported language count Note The official model card says nine languages; the launch article says 35+ languages and 50+ dialects. Both statements are recorded without choosing one. |
| Not published | Model weights and license Note The public API model card does not state downloadable weights or an open-source license. |
| Unresolved | Language-count discrepancy Note The model card says 9 languages, while the current language specification says 35 languages and 50+ dialects. Both values remain visible; the discrepancy is unexplained. |
| Not found in the inspected sources | Artificial Analysis, SpeechArena and TTS Arena exact-model rows Note No exact voxtral-mini-tts-3-0 row was found on these independently checked TTS boards. Artificial Analysis’ generic “Voxtral 3 TTS” label is not tied to the target API ID. This gap is limited to these sources; other source lanes are checked separately. |
| Additional source | Mistral Voxtral 3 announcement (retrieved October 24, 2026) · Mistral Voxtral 3 announcement · Editorial description reviewed October 24, 2026 |
| Additional source | Mistral audio models overview (retrieved October 24, 2026) |
| Additional source | Mistral speech transcription overview (retrieved October 24, 2026) |
| Additional source | Mistral model changelog (retrieved October 24, 2026) |
| Additional source | Mistral inference pricing (retrieved October 24, 2026) |
| Additional source | Mistral text-to-speech overview (retrieved October 24, 2026) · Mistral text-to-speech overview · Editorial description reviewed October 24, 2026 |
| Additional source | Voxtral Mini TTS 3 model card (retrieved October 24, 2026) · Voxtral Mini TTS 3 model card · Editorial description reviewed October 24, 2026 |
| Additional source | VoiceArena US English TTS leaderboard (retrieved October 24, 2026) |
| Additional source | VoiceArena Hindi TTS leaderboard (retrieved October 24, 2026) |
| Additional source | Model metadata (retrieved October 24, 2026) |
| Additional source | Model metadata (retrieved October 24, 2026) |
| Additional source | Model metadata (retrieved October 24, 2026) |
| Additional source | Model metadata (retrieved October 24, 2026) |
| Additional source | Model metadata (retrieved October 24, 2026) |
| Additional source | Model metadata (retrieved October 24, 2026) |
| Additional source | Model metadata (retrieved October 24, 2026) |
| Additional source | Model metadata (retrieved October 24, 2026) |
| Additional source | Model metadata (retrieved October 24, 2026) |
| Additional source | Model metadata (retrieved October 24, 2026) |
| Additional source | Model metadata (retrieved October 24, 2026) |