MI

Mini-Omni2

MiniGGUF by cstrUpdated 2026-08-02
Download on Hugging Face
Parameters782M
Context—
Layers—
Architecturemini-omni2
Smallest quant1.1 GB
Quick answer

Mini-Omni2 needs about 1.9 GB of memory at Q8_0 (a 1.2 GB file) with an 8K context window, so it fits on a 4 GB GPU. This repository has 3 quants, from 1.1 GB (Q4_K) to 1.6 GB (F16).

Mini-Omni2 – a compact multimodal LLM for text, image, audio and voice

Mini-Omni2 is a 0.782 billion‑parameter model that supports multimodal inputs and real‑time speech‑to‑speech. It is released under the MIT licence and is available in several GGUF quantisations. The model can be run locally with llama.cpp or similar inference engines.

Good at

  • Multimodal understanding of image, audio and text
  • Real‑time voice conversation
  • End‑to‑end speech‑to‑speech

Limits

  • No vision support beyond image encoding
  • Requires Whisper for audio encoding
  • No non‑English output

Best for

  • Local multimodal chatbots
  • Voice‑enabled assistants
  • Rapid prototyping of speech‑to‑speech pipelines

Who it’s for. Developers building lightweight multimodal applications who need a small, open‑source model. The model is suitable for local deployment on machines with limited GPU memory.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
Q4_K1.1 GB 1.7 GB
Fits Get file →
Q8_01.2 GB 1.9 GB
Fits Get file →
F161.6 GB 2.3 GB
Fits Get file →

For 8 GB VRAM, use Q4_K (1.06 GB, 1.7 GB memory at 8k context). For 12 GB VRAM, Q8_0 (1.24 GB, 1.9 GB memory). If more memory is available, F16 (1.57 GB, 2.3 GB memory) offers higher precision. The estimates are approximate.

What GPU do you need for Mini-Omni2?

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 F16 (1.6 GB file, about 2.3 GB in use).

Which Mini-Omni2 quant should you choose?

Q8_0 (1.2 GB) is the middle option in this repository. Close to lossless. It costs roughly twice the memory of a 4-bit file, so it suits cases where you have plenty of VRAM.

If memory is tight, the smallest file is Q4_K at 1.1 GB, with the quality trade-offs that come with fewer bits. The usual sweet spot between file size and quality, and the quant most people start with.

New to quant names? Read GGUF quantization explained.

Specifications

Architecture from the GGUF header

Architecturemini-omni2
Parameters782M

Source

Quantized bycstr
FamilyMini
Licensemit
Downloads (HF)1,786
Last updated2026-08-02

Frequently asked questions about Mini-Omni2

How much VRAM does Mini-Omni2 need?

About 1.9 GB at Q8_0 with an 8K context window. That is the 1.2 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 Mini-Omni2 on an 8 GB GPU?

Yes, with F16: the 1.6 GB file needs about 2.3 GB at 8K context. Larger quants in this repository do not fit in 8 GB.

What is the file size of Mini-Omni2 GGUF?

It depends on the quant: Q4_K is 1.1 GB, Q8_0 is 1.2 GB, F16 is 1.6 GB.

Where can I download Mini-Omni2 GGUF?

The files are on Hugging Face in the cstr/mini-omni2-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 Mini-Omni2 use?

Hugging Face lists the license as mit. 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 Mini‑Omni2 on 8 GB VRAM?

Yes, the Q4_K quantisation requires about 1.7 GB of VRAM at 8k context, which fits within 8 GB.

Does Mini‑Omni2 support languages other than English?

The model is trained only on English, but its Whisper encoder can understand other languages; output remains in English.

What is required to run the speech‑to‑speech feature?

The SNAC 24kHz codec companion is needed for TTS and speech‑to‑speech; it must be installed alongside the model.

Similar models

Browse all →

More from cstr

All by cstr →

More Mini models

All Mini →

Other tiny models

All tiny models →