Mistral AILanguage model
Voxtral Small (latest)
Voxtral Small (latest) has no overall index value yet because there are too few independent benchmark results.
19out of 100+3 vs. last month
based on two data groups, mostly search
Summary
Voxtral Small (latest) is the rolling API entry for Mistral’s 24B audio-language model. It accepts audio and text and supports transcription, questions about recordings, and summaries. The alias corresponds to the fixed checkpoint Voxtral Small 24B.
The alias can change while the fixed checkpoint’s model card remains unchanged.
All benchmark results
Every published result with its source, test setup and comparison models.
Show all 3 published benchmark results
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.
speech-emotion-bench IEMOCAP three-class valence accuracy (Show measurement, test conditions, and source)
- Source value
- 54.4
- Score
- 54.4
- Metric
- accuracy
- Unit
- %
- Benchmark version
- speech-emotion-bench August 2026 snapshot, evaluated 2026-08-01
- Category
- audio
- Direction
- Higher is better
- Sample
- 1947
- Confidence interval
- 2.2
- Source type
- vendor-reported
- Evaluator
- oruk speech-emotion-bench
- Status
- active
- Retrieved at
- 2026-10-16
- Methodology
- Vendor-reported result. The report says identical audio from the 2,008-clip IEMOCAP test split was sent to each vendor public API in one harness. Three-class valence mapped angry/frustrated/sad to negative, neutral to neutral, and happy/excited to positive. Exact endpoint label: voxtral-small-latest. 1,947 scored clips; reported ± is half the 95% Wilson interval.
- Context
- This is a dated, prompted emotion-classification result on acted IEMOCAP test audio, not a general transcription or conversational-audio score. Oruk sells the competing Resonance model shown in the same IEMOCAP report, creating a direct commercial conflict; this is vendor-reported, not independent third-party validation. The endpoint is a mutable latest alias; the tested backend revision is not published and is not equated to fixed voxtral-small-2507.
speech-emotion-bench IEMOCAP four-class accuracy (Show measurement, test conditions, and source)
- Source value
- 44
- Score
- 44
- Metric
- accuracy
- Unit
- %
- Benchmark version
- speech-emotion-bench August 2026 snapshot, evaluated 2026-08-01
- Category
- audio
- Direction
- Higher is better
- Sample
- 1384
- Source type
- vendor-reported
- Evaluator
- oruk speech-emotion-bench
- Status
- active
- Retrieved at
- 2026-10-16
- Methodology
- Vendor-reported result. Four acoustic classes: angry, happy+excited, sad, neutral. Predictions outside those four classes count as errors. Exact endpoint label in the related IEMOCAP API table: voxtral-small-latest. n=1,384. The source prints 44.0 ± 2.6%; the interval method for accuracy is not specified in this section.
- Context
- This is a separate four-class score from the report's three-class valence result. Do not compare the two tasks. Oruk sells the competing Resonance model shown in the same IEMOCAP report, creating a direct commercial conflict; this is vendor-reported, not independent third-party validation. The source does not state the accuracy interval method here; the underlying latest alias revision is not pinned.
speech-emotion-bench IEMOCAP four-class unweighted average recall (Show measurement, test conditions, and source)
- Source value
- 44.7
- Score
- 44.7
- Metric
- unweighted average recall
- Unit
- %
- Benchmark version
- speech-emotion-bench August 2026 snapshot, evaluated 2026-08-01
- Category
- audio
- Direction
- Higher is better
- Sample
- 1384
- Source type
- vendor-reported
- Evaluator
- oruk speech-emotion-bench
- Status
- active
- Retrieved at
- 2026-10-16
- Methodology
- Vendor-reported result. Four acoustic classes: angry, happy+excited, sad, neutral; predictions outside these four classes count as errors; UAR weights each class equally and its intervals are bootstrap. Exact endpoint label in the related IEMOCAP API table: voxtral-small-latest. n=1,384.
- Context
- This four-class UAR is not comparable to the report's three-class valence accuracy or four-class accuracy. Oruk sells the competing Resonance model shown in the same IEMOCAP report, creating a direct commercial conflict; this is vendor-reported, not independent third-party validation. The source does not publish an immutable backend revision for the latest alias.
Popularity compared
How often the model is searched for, used and mentioned, compared with every model of its category.
Popularity index
The 100 most popular models of this category, 0 to 100, higher is more popular
20 of 101 models
As of November 2026 · Source: gradually.ai popularity index
Over time
Weekly values from October 18, 2026 to November 7, 2026
Voxtral Small (latest)Codestral-22B-v0.1 19DBRX 19Llama 3.1 70B 19Step 3.5 Flash 2603 19
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.
Similar models
More models of the same kind as Voxtral Small (latest), chosen by model line, developer, size, price and release date.
- Magistral SmallMistral AI · same developer
- Devstral SmallMistral AI · same developer
- Voxtral Small 24BMistral AI · same developer
- Magistral Small 1.2Mistral AI · same developer
- Devstral Small 2Mistral AI · same developer
- Ministral 3 14BMistral AI · same developer
- Codestral-22B-v0.1Mistral AI · same developer
- Codestral (latest)Mistral AI · same developer
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
Frequently asked questions about Voxtral Small (latest)
How large is the context window of Voxtral Small (latest)?
32,000 tokens. That is how much text Voxtral Small (latest) reads at once, about 50 A4 pages.
Page count as a rule of thumb from Artificial Analysis
Can I run Voxtral Small (latest) locally?
Yes, the weights are open.
Stated by Mistral AI
- Open weights
- yes
- License
- Apache 2.0
- Parameters
- 24B
- active per token
- n/a
Sources and data date
Show all 15 sources
Benchmark sources at a glance
Which independent evaluators tested this model. This is coverage, not a quality ranking.
No active result from the reviewed external evaluators can currently be attributed to this model. Other source evidence remains available below.
Without vendor claims and historical or disputed results; those stay labelled in the single values.
| Type | Evidence and data date |
|---|---|
| Input | Documented modalities. Retrieved October 27, 2026. |
| Output | Documented modalities. Retrieved October 27, 2026. |
| Research status | Research date October 15, 2026. 7 source URLs checked. This documents the inspected sources, not an exhaustive inventory of every publication. |
| Unresolved | immutable-endpoint-checkpoint Note The evaluator names the exact public API alias voxtral-small-latest, so its dated alias-level results are recorded. The underlying backend snapshot is not disclosed; Mistral documents latest aliases as potentially silently updated. No equivalence to the fixed voxtral-small-2507 checkpoint is claimed. |
| Additional source | Hugging Face (retrieved October 15, 2026) · Voxtral Small 24B model card · Editorial description reviewed October 2, 2026 |
| Additional source | oruk speech-emotion-bench (retrieved October 16, 2026) |
| Additional source | Model metadata (retrieved October 16, 2026)1234 |
| Additional source | Voxtral announcement · Editorial description reviewed October 2, 2026 |
| Additional source | Mistral model overview · Editorial description reviewed October 2, 2026 |
| 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 |