Modelofficial release

mistralai/Voxtral-Mini-4B-Realtime-2602

Mistral models →Hugging Face ↗
Parameters4.4B
Context128K
Input → outputaudio → text
Licenseapache-2.0
PrecisionBF16
ArchitectureVoxtralRealtimeForConditionalGeneration
ReleasedJan 21, 2026
Downloads1.9M

Based on mistralai/Ministral-3-3B-Base-2512

Other sizes and formats in the family

ModelParamsPrecisionDownloads
mistralai/Voxtral-Mini-4B-Realtime-Arabic4.4BBF16123

Compare side by side →

Built on this model

ModelParamsPrecisionDownloads
mistralai/Voxtral-Mini-4B-Realtime-Arabic4.4BBF16123

This release in the news

No coverage of this exact release yet.

Recent news about Mistral models

  1. Oct 11, 2026 · Opinion / analysis
    Every Model That Can Be Run On 10-16GB VRAM Ranked
  2. Oct 11, 2026 · Opinion / analysis
    Local Voice Assistant based on Qwen3.5 4B with skills/toolcalling. No dedicated GPU. Local STT+TTS
  3. Oct 10, 2026 · Research paper
    Sharing AI progress in mathematics
  4. Oct 8, 2026 · Model release
    mistralai/Voxtral-Mini-4B-Realtime-Arabic
  5. Oct 7, 2026 · Product / feature launch
    Introducing Mistral Large 4
  6. Oct 7, 2026 · Research paper
    Activation-Aware Weight Tensorization: A Calibration-Time Preconditioner for Tensor-Network LLM Compression
  7. Oct 7, 2026 · Open-source release
    browser-use/browser-use 0.13.11
  8. Oct 6, 2026 · Model release
    llm-mistral 0.16