OmniASR-LLM-300M needs about 2.6 GB of memory at Q8_0 (a 1.8 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 3.3 GB (F16).
OmniASR-LLM-300M: 1.63B parameter multilingual ASR model
OmniASR-LLM-300M is a 1.63B parameter end‑to‑end automatic speech recognition model. It supports 1600+ languages and is released under Apache‑2.0. The GGUF conversion is available for llama.cpp inference.
Good at
- Multilingual speech recognition
- Large language coverage
- Low‑parameter size
Limits
- No vision support
- Requires libsndfile for audio
- No mixture of experts
Best for
- Embedded ASR on edge devices
- Low‑latency transcription
- Multilingual voice assistants
Who it’s for. Developers building speech‑to‑text applications that need broad language support in a compact model. Ideal for projects with limited GPU memory and Apache‑2.0 licensing requirements.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| Q4_K | 1.1 GB | 1.7 GB | Fits | Get file → | |
| Q8_0 | 1.8 GB | 2.6 GB | Fits | Get file → | |
| F16 | 3.3 GB | 4.1 GB | Fits | Get file → |
For 8 GB VRAM, use Q4_K (1.07 GB file, 1.7 GB memory at 8k context). For 12 GB VRAM, Q8_0 (1.84 GB file, 2.6 GB memory). F16 (3.26 GB file, 4.1 GB memory) requires more memory but offers higher precision.
What GPU do you need for OmniASR-LLM-300M?
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 Q8_0 (1.8 GB file, about 2.6 GB in use).
- 6 GB VRAM: the best quant that fits is F16 (3.3 GB file, about 4.1 GB in use).
Which OmniASR-LLM-300M quant should you choose?
Q8_0 (1.8 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
Source
Frequently asked questions about OmniASR-LLM-300M
How much VRAM does OmniASR-LLM-300M need?
About 2.6 GB at Q8_0 with an 8K context window. That is the 1.8 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 OmniASR-LLM-300M on an 8 GB GPU?
Yes, with F16: the 3.3 GB file needs about 4.1 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of OmniASR-LLM-300M GGUF?
It depends on the quant: Q4_K is 1.1 GB, Q8_0 is 1.8 GB, F16 is 3.3 GB.
Where can I download OmniASR-LLM-300M GGUF?
The files are on Hugging Face in the cstr/omniasr-llm-300m-v2-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 OmniASR-LLM-300M 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 OmniASR-LLM-300M on 8 GB VRAM?
Yes, the Q4_K quantized model is 1.07 GB and requires approximately 1.7 GB of VRAM at 8k context.
Does this model support vision tasks?
No, OmniASR-LLM-300M is an audio-only ASR model.
What license applies to the GGUF files?
The GGUF files are distributed under Apache‑2.0, the same license as the original model.