Voxtral Mini 4B Realtime needs about 3.7 GB of memory at Q4_K_M (a 2.8 GB file) with an 8K context window, so it fits on a 4 GB GPU. This repository has 6 quants, from 2.8 GB (Q4_K_M) to 8.9 GB (F16).
Voxtral Mini 4B Realtime – real‑time multilingual speech‑to‑text
Voxtral Mini 4B Realtime is a 4.8 B parameter model designed for low‑latency audio transcription. It streams audio with a 480 ms delay and supports up to 13 languages. The model is released under Apache‑2.0 and is compatible with vLLM’s Realtime API.
Good at
- Real‑time speech transcription
- Low‑latency audio streaming
- Multilingual support
Limits
- Requires vLLM for inference
- No vision or image capabilities
- Maximum context length limited by RoPE pre‑allocations
Best for
- Live captioning of meetings
- Real‑time podcast transcription
- Audio‑to‑text for accessibility
Who it’s for. Developers building real‑time transcription services or accessibility tools. The model is suitable for applications that need low latency and multilingual coverage without visual input.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| Q4_K_M | 2.8 GB | 3.7 GB | Fits | Get file → | |
| Q5_K_M | 3.3 GB | 4.2 GB | Fits | Get file → | |
| Q6_K | 3.7 GB | 4.6 GB | Fits | Get file → | |
| Q8_0 | 4.7 GB | 5.8 GB | Fits | Get file → | |
| BF16 | 8.9 GB | 10.4 GB | Too big | Get file → | |
| F16 | 8.9 GB | 10.4 GB | Too big | Get file → |
For 8 GB VRAM, use Q4_K_M (2.83 GB file, ~3.7 GB VRAM at 8k context). For 12 GB VRAM, Q5_K_M (3.28 GB file, ~4.2 GB VRAM). For 16 GB VRAM, Q6_K (3.66 GB file, ~4.6 GB VRAM). Memory estimates are approximate.
What GPU do you need for Voxtral Mini 4B Realtime?
These figures add the file size, an fp16 KV cache at an 8K context window and about 0.5 GB for the runtime. They are estimates, so leave a little headroom, and expect to use less context or a smaller quant if you are right at the limit.
- 4 GB VRAM: the best quant that fits is Q4_K_M (2.8 GB file, about 3.7 GB in use).
- 6 GB VRAM: the best quant that fits is Q8_0 (4.7 GB file, about 5.8 GB in use).
- 8 GB VRAM: the best quant that fits is Q8_0 (4.7 GB file, about 5.8 GB in use).
- 12 GB VRAM: the best quant that fits is F16 (8.9 GB file, about 10.4 GB in use).
Which Voxtral Mini 4B Realtime quant should you choose?
Q4_K_M is the usual starting point for Voxtral Mini 4B Realtime: it balances file size and quality well. Here it is a 2.8 GB file.
The Q8_0 file is 4.7 GB, about 1.7 times the size of Q4_K_M. It keeps the weights closer to the original, which only matters if you have the memory to spare.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about Voxtral Mini 4B Realtime
How much VRAM does Voxtral Mini 4B Realtime need?
About 3.7 GB at Q4_K_M with an 8K context window. That is the 2.8 GB file, plus the KV cache and roughly 0.5 GB of runtime overhead. Smaller quants need less and longer context needs more.
Can I run Voxtral Mini 4B Realtime on an 8 GB GPU?
Yes, with Q8_0: the 4.7 GB file needs about 5.8 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of Voxtral Mini 4B Realtime GGUF?
It depends on the quant: Q4_K_M is 2.8 GB, Q5_K_M is 3.3 GB, Q6_K is 3.7 GB, Q8_0 is 4.7 GB, BF16 is 8.9 GB, F16 is 8.9 GB.
Where can I download Voxtral Mini 4B Realtime GGUF?
The files are on Hugging Face in the handy-computer/Voxtral-Mini-4B-Realtime-2602-gguf repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.
What license does Voxtral Mini 4B Realtime use?
Hugging Face lists the license as apache-2.0. Licenses can differ between the quantized files and the original model, so check the repository and the base model before commercial use.
Can I run Voxtral Mini 4B Realtime on 8 GB VRAM?
Yes, using the Q4_K_M quantization which requires about 3.7 GB VRAM at 8k context.
Does the model support image input?
No, it is a speech‑to‑text model with no vision capabilities.
What is the maximum transcription length?
The default vLLM configuration sets a maximum model length of 131072 tokens, roughly 3 hours of audio.