Mistral AIAudio modelActive
Voxtral Mini Transcribe 2
PopularityRank 524 of 876
26out of 100same as last month
Summary
Voxtral Mini Transcribe 2 is Mistral’s optimized transcription service for audio files. The API alias voxtral-mini-latest supports timestamps, speaker diarization, and context terms. With context_bias, you can supply up to 100 words or phrases to improve names and domain vocabulary. The documentation says context biasing is optimized for English, while support for other languages remains experimental.
Voxtral Mini Transcribe 2 processes audio files with timestamps, speaker separation, and context terms. Mistral supports streaming in the transcription service, but documents Voxtral Mini Transcribe Realtime as the separate live path. Batch and real time therefore remain distinct model paths.
All benchmark results
Every published result with its source, test setup and comparison models.
Show all 56 published benchmark results
Page 1 of 5
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.
Artificial Analysis AA-WER (non-streaming) (Show measurement, test conditions, and source)
- Source value
- 3.6
- Score
- 3.6
- Metric
- word error rate (WER)
- Unit
- %
- Benchmark version
- AA-WER v2.2
- Category
- audio
- Direction
- Lower is better
- Source type
- independent-evaluator
- Evaluator
- Artificial Analysis
- Status
- active
- Retrieved at
- 2026-10-15
- Methodology
- Independent evaluation. Artificial Analysis directly tested the named API endpoint; audio-duration-weighted WER across AA-AgentTalk (50%), VoxPopuli-Cleaned-AA English (25%), and Earnings22-Cleaned-AA (25%), about eight hours total. Current normalizer AA-WER v2.2.
Mistral-reported approximate FLEURS word error rate (Voxtral Mini Transcribe V2) (Show measurement, test conditions, and source)
- Source value
- 4
- Score
- 4
- Metric
- Approximately reported word error rate on FLEURS
- Unit
- %
- Benchmark version
- Mistral Voxtral Transcribe 2 announcement, approximate FLEURS claim, checked 2026-10-16
- Category
- audio
- Direction
- Lower is better
- Source type
- vendor-reported
- Evaluator
- Mistral
- Status
- active
- Retrieved at
- 2026-10-16
- Methodology
- Vendor-reported result. Mistral's announcement explicitly attributes an approximately 4% FLEURS WER to Voxtral Mini Transcribe V2. The numeric score 4 is only a representation of that approximate claim, not an exact or unrounded result. Mistral does not publish the sample denominator, confidence interval, or full evaluation configuration here.
- Context
- This is an explicitly approximate vendor claim, not an exact measured value. The source does not provide the evaluation sample denominator, confidence interval, or enough protocol detail to establish comparability with other FLEURS results. Do not interpret 4% as precise or compare it directly with Artificial Analysis AA-WER.
Voxtral Mini Transcribe V2, Voxtral Realtime paper Table 3 English short-form macro-average WER (Show measurement, test conditions, and source)
- Source value
- 7.27
- Score
- 7.27
- Metric
- word error rate (WER)
- Unit
- %
- Benchmark version
- Voxtral Realtime arXiv:2602.11298v3, Table 3
- Category
- audio
- Direction
- Lower is better
- Source type
- vendor-reported
- Evaluator
- Mistral AI, Voxtral Realtime paper
- Status
- active
- Retrieved at
- 2026-10-16
- Methodology
- Vendor-reported result. Exact paper result row "Voxtral Mini Transcribe V2" mapped by name to local profile "voxtral-mini-transcribe-2" (Voxtral Mini Transcribe 2); English short-form macro-average WER. The paper does not specify an immutable served API snapshot or per-metric evaluated sample count. Table 3 macro-averages English results across tasks.
- Context
- Mistral-authored comparison; the paper identifies the row as Voxtral Mini Transcribe V2, but does not state an immutable served API snapshot. The paper does not publish a metric-specific evaluated sample count or confidence interval.
Voxtral Mini Transcribe V2, Voxtral Realtime paper Table 3 English long-form macro-average WER (Show measurement, test conditions, and source)
- Source value
- 7.11
- Score
- 7.11
- Metric
- word error rate (WER)
- Unit
- %
- Benchmark version
- Voxtral Realtime arXiv:2602.11298v3, Table 3
- Category
- audio
- Direction
- Lower is better
- Source type
- vendor-reported
- Evaluator
- Mistral AI, Voxtral Realtime paper
- Status
- active
- Retrieved at
- 2026-10-16
- Methodology
- Vendor-reported result. Exact paper result row "Voxtral Mini Transcribe V2" mapped by name to local profile "voxtral-mini-transcribe-2" (Voxtral Mini Transcribe 2); English long-form macro-average WER. The paper does not specify an immutable served API snapshot or per-metric evaluated sample count. Table 3 macro-averages English results across tasks.
- Context
- Mistral-authored comparison; the paper identifies the row as Voxtral Mini Transcribe V2, but does not state an immutable served API snapshot. The paper does not publish a metric-specific evaluated sample count or confidence interval.
Voxtral Mini Transcribe V2, Voxtral Realtime paper Table 3 FLEURS macro-average error rate (Show measurement, test conditions, and source)
- Source value
- 5.9
- Score
- 5.9
- Metric
- macro-average error rate (WER except CER for Japanese and Chinese)
- Unit
- %
- Benchmark version
- Voxtral Realtime arXiv:2602.11298v3, Table 3
- Category
- audio
- Direction
- Lower is better
- Source type
- vendor-reported
- Evaluator
- Mistral AI, Voxtral Realtime paper
- Status
- active
- Retrieved at
- 2026-10-16
- Methodology
- Vendor-reported result. Exact paper result row "Voxtral Mini Transcribe V2" mapped by name to local profile "voxtral-mini-transcribe-2" (Voxtral Mini Transcribe 2); FLEURS macro-average error rate. The paper does not specify an immutable served API snapshot or per-metric evaluated sample count. Table 3 averages across languages; Japanese and Chinese use CER, while other languages use WER.
- Context
- Mistral-authored comparison; the paper identifies the row as Voxtral Mini Transcribe V2, but does not state an immutable served API snapshot. The paper does not publish a metric-specific evaluated sample count or confidence interval.
Voxtral Mini Transcribe V2, Voxtral Realtime paper Table 3 Mozilla Common Voice macro-average error rate (Show measurement, test conditions, and source)
- Source value
- 8.07
- Score
- 8.07
- Metric
- macro-average error rate (WER except CER for Japanese and Chinese; Arabic excluded from this macro-average)
- Unit
- %
- Benchmark version
- Voxtral Realtime arXiv:2602.11298v3, Table 3
- Category
- audio
- Direction
- Lower is better
- Source type
- vendor-reported
- Evaluator
- Mistral AI, Voxtral Realtime paper
- Status
- active
- Retrieved at
- 2026-10-16
- Methodology
- Vendor-reported result. Exact paper result row "Voxtral Mini Transcribe V2" mapped by name to local profile "voxtral-mini-transcribe-2" (Voxtral Mini Transcribe 2); Mozilla Common Voice macro-average error rate. The paper does not specify an immutable served API snapshot or per-metric evaluated sample count. Table 3 averages across languages and excludes Arabic from the MCV macro-average; Japanese and Chinese use CER, while other languages use WER.
- Context
- Mistral-authored comparison; the paper identifies the row as Voxtral Mini Transcribe V2, but does not state an immutable served API snapshot. The paper does not publish a metric-specific evaluated sample count or confidence interval.
Voxtral Mini Transcribe V2, Voxtral Realtime paper Appendix Table 5 English short-form WER: LibriSpeech Test Clean (LS-C) (Show measurement, test conditions, and source)
- Source value
- 1.6
- Score
- 1.6
- Metric
- word error rate (WER)
- Unit
- %
- Benchmark version
- Voxtral Realtime arXiv:2602.11298v3, Appendix Table 5
- Category
- audio
- Direction
- Lower is better
- Source type
- vendor-reported
- Evaluator
- Mistral AI, Voxtral Realtime paper
- Status
- active
- Retrieved at
- 2026-10-16
- Methodology
- Vendor-reported result. Exact paper result row "Voxtral Mini Transcribe V2" mapped by name to local profile "voxtral-mini-transcribe-2" (Voxtral Mini Transcribe 2); English short-form WER: LibriSpeech Test Clean (LS-C). The paper does not specify an immutable served API snapshot or per-metric evaluated sample count. Table 5 reports dataset-level WER; its AVG cell is omitted here because the same macro-average appears in Table 3.
- Context
- Mistral-authored comparison; the paper identifies the row as Voxtral Mini Transcribe V2, but does not state an immutable served API snapshot. The paper does not publish a metric-specific evaluated sample count or confidence interval.
Voxtral Mini Transcribe V2, Voxtral Realtime paper Appendix Table 5 English short-form WER: LibriSpeech Test Other (LS-O) (Show measurement, test conditions, and source)
- Source value
- 3.24
- Score
- 3.24
- Metric
- word error rate (WER)
- Unit
- %
- Benchmark version
- Voxtral Realtime arXiv:2602.11298v3, Appendix Table 5
- Category
- audio
- Direction
- Lower is better
- Source type
- vendor-reported
- Evaluator
- Mistral AI, Voxtral Realtime paper
- Status
- active
- Retrieved at
- 2026-10-16
- Methodology
- Vendor-reported result. Exact paper result row "Voxtral Mini Transcribe V2" mapped by name to local profile "voxtral-mini-transcribe-2" (Voxtral Mini Transcribe 2); English short-form WER: LibriSpeech Test Other (LS-O). The paper does not specify an immutable served API snapshot or per-metric evaluated sample count. Table 5 reports dataset-level WER; its AVG cell is omitted here because the same macro-average appears in Table 3.
- Context
- Mistral-authored comparison; the paper identifies the row as Voxtral Mini Transcribe V2, but does not state an immutable served API snapshot. The paper does not publish a metric-specific evaluated sample count or confidence interval.
Voxtral Mini Transcribe V2, Voxtral Realtime paper Appendix Table 5 English short-form WER: GigaSpeech (GS) (Show measurement, test conditions, and source)
- Source value
- 10.39
- Score
- 10.39
- Metric
- word error rate (WER)
- Unit
- %
- Benchmark version
- Voxtral Realtime arXiv:2602.11298v3, Appendix Table 5
- Category
- audio
- Direction
- Lower is better
- Source type
- vendor-reported
- Evaluator
- Mistral AI, Voxtral Realtime paper
- Status
- active
- Retrieved at
- 2026-10-16
- Methodology
- Vendor-reported result. Exact paper result row "Voxtral Mini Transcribe V2" mapped by name to local profile "voxtral-mini-transcribe-2" (Voxtral Mini Transcribe 2); English short-form WER: GigaSpeech (GS). The paper does not specify an immutable served API snapshot or per-metric evaluated sample count. Table 5 reports dataset-level WER; its AVG cell is omitted here because the same macro-average appears in Table 3.
- Context
- Mistral-authored comparison; the paper identifies the row as Voxtral Mini Transcribe V2, but does not state an immutable served API snapshot. The paper does not publish a metric-specific evaluated sample count or confidence interval.
Voxtral Mini Transcribe V2, Voxtral Realtime paper Appendix Table 5 English short-form WER: VoxPopuli (VP) (Show measurement, test conditions, and source)
- Source value
- 6.81
- Score
- 6.81
- Metric
- word error rate (WER)
- Unit
- %
- Benchmark version
- Voxtral Realtime arXiv:2602.11298v3, Appendix Table 5
- Category
- audio
- Direction
- Lower is better
- Source type
- vendor-reported
- Evaluator
- Mistral AI, Voxtral Realtime paper
- Status
- active
- Retrieved at
- 2026-10-16
- Methodology
- Vendor-reported result. Exact paper result row "Voxtral Mini Transcribe V2" mapped by name to local profile "voxtral-mini-transcribe-2" (Voxtral Mini Transcribe 2); English short-form WER: VoxPopuli (VP). The paper does not specify an immutable served API snapshot or per-metric evaluated sample count. Table 5 reports dataset-level WER; its AVG cell is omitted here because the same macro-average appears in Table 3.
- Context
- Mistral-authored comparison; the paper identifies the row as Voxtral Mini Transcribe V2, but does not state an immutable served API snapshot. The paper does not publish a metric-specific evaluated sample count or confidence interval.
Voxtral Mini Transcribe V2, Voxtral Realtime paper Appendix Table 5 English short-form WER: SwitchBoard (SB) (Show measurement, test conditions, and source)
- Source value
- 11.54
- Score
- 11.54
- Metric
- word error rate (WER)
- Unit
- %
- Benchmark version
- Voxtral Realtime arXiv:2602.11298v3, Appendix Table 5
- Category
- audio
- Direction
- Lower is better
- Source type
- vendor-reported
- Evaluator
- Mistral AI, Voxtral Realtime paper
- Status
- active
- Retrieved at
- 2026-10-16
- Methodology
- Vendor-reported result. Exact paper result row "Voxtral Mini Transcribe V2" mapped by name to local profile "voxtral-mini-transcribe-2" (Voxtral Mini Transcribe 2); English short-form WER: SwitchBoard (SB). The paper does not specify an immutable served API snapshot or per-metric evaluated sample count. Table 5 reports dataset-level WER; its AVG cell is omitted here because the same macro-average appears in Table 3.
- Context
- Mistral-authored comparison; the paper identifies the row as Voxtral Mini Transcribe V2, but does not state an immutable served API snapshot. The paper does not publish a metric-specific evaluated sample count or confidence interval.
Voxtral Mini Transcribe V2, Voxtral Realtime paper Appendix Table 5 English short-form WER: CallHome (CH) (Show measurement, test conditions, and source)
- Source value
- 12.74
- Score
- 12.74
- Metric
- word error rate (WER)
- Unit
- %
- Benchmark version
- Voxtral Realtime arXiv:2602.11298v3, Appendix Table 5
- Category
- audio
- Direction
- Lower is better
- Source type
- vendor-reported
- Evaluator
- Mistral AI, Voxtral Realtime paper
- Status
- active
- Retrieved at
- 2026-10-16
- Methodology
- Vendor-reported result. Exact paper result row "Voxtral Mini Transcribe V2" mapped by name to local profile "voxtral-mini-transcribe-2" (Voxtral Mini Transcribe 2); English short-form WER: CallHome (CH). The paper does not specify an immutable served API snapshot or per-metric evaluated sample count. Table 5 reports dataset-level WER; its AVG cell is omitted here because the same macro-average appears in Table 3.
- Context
- Mistral-authored comparison; the paper identifies the row as Voxtral Mini Transcribe V2, but does not state an immutable served API snapshot. The paper does not publish a metric-specific evaluated sample count or confidence interval.
Page 1 of 5
Popularity compared
How often the model is searched for, used and mentioned, compared with every model of its category.
Popularity index
0 to 100, higher is more popular
20 of 83 models
As of November 2026 · Source: gradually.ai popularity index
Over time
Weekly values from October 18, 2026 to November 6, 2026
Voxtral Mini Transcribe 2Scribe v2 Realtime 29Lyria 3 Clip 26Index-Echo-S2TT-2B 26Index-Echo-S2TT-9B 24
Method changes in the timeline, a jump at these points is no market event:
- 1November 1, 2026: Value only with search or usage data, 30-day attention
What is gaining popularity right now, across models, vendors, harnesses and AI subscriptions, is collected in All AI trends.
Voxtral Mini Transcribe 2 compared
- 1Voxtral Mini Transcribe 2 vs. Gemini 3.5 Transcribe1 shared test
- 2Voxtral Mini Transcribe 2 vs. GPT-4o Mini Transcribe1 shared test
- 3Voxtral Mini Transcribe 2 vs. GPT-Transcribe1 shared test
- 4Voxtral Mini Transcribe 2 vs. Scribe v21 shared test
- 5Voxtral Mini Transcribe 2 vs. Nova-31 shared test
- 6Voxtral Mini Transcribe 2 vs. MAI-Transcribe-21 shared test
- 7Voxtral Mini Transcribe 2 vs. Muse Voice Transcribe1 shared test
Similar models
More models of the same kind as Voxtral Mini Transcribe 2, chosen by model line, developer, size, price and release date.
- Voxtral Mini TTSMistral AI · same developer
- Voxtral Mini Transcribe RealtimeMistral AI · same developer
- Voxtral Mini Transcribe 3Mistral AI · same developer
- Voxtral Mini Transcribe Realtime 3Mistral AI · same developer
- Voxtral Mini TTS 3Mistral AI · same developer
- Voxtral Mini TTS Realtime 3Mistral AI · same developer
- Voxtral Small Transcribe 3Mistral AI · same developer
- Gemini 3.5 TranscribeGoogle · popular in this category
All specifications
Architecture
Input and output
Watermarking and provenance
AI watermarking refers to invisible signals embedded in the output. Visible marks and provenance metadata are listed separately. Unknown means there is insufficient evidence for that particular field.
- AI watermarking
- Unknown
- no source
- Watermarking technique
- Not known
- no source
Licence and access
API
Further details
Sources and data date
Show all 20 sources
Benchmark sources at a glance
Which independent evaluators tested this model. This is coverage, not a quality ranking.
| Evaluator | Test series | Composite indices | Recorded active measurements |
|---|---|---|---|
| Artificial Analysis | 0 | 0 | 1 |
Without vendor claims and historical or disputed results; those stay labelled in the single values. 1 measurements are not yet assigned to a test series.
| Type | Evidence and data date |
|---|---|
| Research status | Research date October 15, 2026. 15 source URLs checked. This documents the inspected sources, not an exhaustive inventory of every publication. |
| Not published | fleurs-evaluation-denominator-protocol-and-confidence-interval Note Mistral publishes an approximate model-specific 4% FLEURS WER claim, recorded above as approximate. The exact unrounded score, sample denominator, confidence interval, and full evaluation protocol are not published in the cited announcement. |
| Not published | immutable-evaluated-api-snapshot Note The paper names Voxtral Mini Transcribe V2 but does not disclose the immutable API/model snapshot evaluated; the exact product name is the available profile bridge. |
| Not published | evaluated-denominator-and-confidence-interval Note The paper tables do not publish the metric-specific evaluated sample count or confidence intervals; no denominator was inferred from corpus sizes. |
| Not published | vals-api-serving-revision Note Neither Omi nor Vals names an immutable serving revision for the Voxtral API call. The evaluators test different datasets and tasks. |
| Additional source | Mistral audio documentation (retrieved September 22, 2026; October 15, 2026) · Mistral AI offline transcription documentation · Editorial description reviewed October 2, 2026 |
| Additional source | Artificial Analysis (retrieved October 15, 2026) |
| Additional source | Mistral (retrieved October 16, 2026) |
| Additional source | Mistral AI, Voxtral Realtime paper (retrieved October 16, 2026) |
| Additional source | Omi Health Medical STT Benchmark (retrieved October 17, 2026)123 |
| Additional source | Vals AI VoiceCodeBench (retrieved October 17, 2026) |
| Additional source | Ocular AI, Converse-STT (retrieved October 17, 2026) |
| Additional source | David AI Research, DAI-ASR-I18N (retrieved October 17, 2026) |
| Additional source | Model metadata (retrieved October 15, 2026) |
| Additional source | Model metadata (retrieved October 16, 2026) |
| Additional source | Model metadata (retrieved October 17, 2026)12 |
| Additional source | OpenRouter: Popularity index (derived), CC BY 4.0 |
| Additional source | Vercel AI Gateway: Popularity index (derived), CC BY 4.0 |
| Additional source | Ollama: cumulative downloads for the popularity index |
| Additional source | Replicate: cumulative runs for the popularity index |
| Additional source | Google Trending Now: limited feed selection for Germany and the US |
| Additional source | Creative Commons Attribution 4.0 |