Breeze-ASR-25 needs about 1.6 GB of memory at Q4_K_M (a 1.0 GB file) with an 8K context window, so it fits on a 4 GB GPU. This repository has 6 quants, from 1.0 GB (Q4_K_M) to 3.1 GB (F16).
Breeze-ASR-25: Whisper-large-v2 fine‑tuned for Taiwanese Mandarin and Mandarin‑English code‑switching
Breeze-ASR-25 is a 1.5‑billion‑parameter Whisper‑based ASR model. It is optimized for Traditional Chinese and English, with strong support for intra‑ and inter‑sentential code‑switching. The model includes enhanced timestamp alignment for automatic captioning.
Good at
- Traditional Chinese transcription
- Mandarin‑English code‑switching
- Timestamp‑aligned captions
Limits
- No vision support
- Requires at least 8‑k context for full performance
- Apache‑2.0 license
Best for
- Transcribing Taiwanese Mandarin podcasts
- Generating bilingual subtitles
- Processing long‑form audio with code‑switching
Who it’s for. Developers and researchers needing accurate transcription of Taiwanese Mandarin and Mandarin‑English mixed speech. The model is suitable for applications such as subtitle generation, voice‑to‑text services, and linguistic research on code‑switching.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| Q4_K_M | 1.0 GB | 1.6 GB | Fits | Get file → | |
| Q5_K_M | 1.2 GB | 1.8 GB | Fits | Get file → | |
| Q6_K | 1.3 GB | 2.0 GB | Fits | Get file → | |
| Q8_0 | 1.7 GB | 2.4 GB | Fits | Get file → | |
| BF16 | 3.1 GB | 4.0 GB | Fits | Get file → | |
| F16 | 3.1 GB | 4.0 GB | Fits | Get file → |
For 8 GB VRAM, Q4_K_M (1 GB) or Q5_K_M (1.16 GB) are suitable, using about 1.6–1.8 GB of VRAM at 8 k context. If more memory is available, Q6_K (1.3 GB) or Q8_0 (1.67 GB) provide slightly higher precision. BF16 and F16 (≈3.1 GB) require more than 8 GB VRAM.
What GPU do you need for Breeze-ASR-25?
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 (3.1 GB file, about 4.0 GB in use).
Which Breeze-ASR-25 quant should you choose?
Q4_K_M is the usual starting point for Breeze-ASR-25: it balances file size and quality well. Here it is a 1.0 GB file.
The Q8_0 file is 1.7 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 Breeze-ASR-25
How much VRAM does Breeze-ASR-25 need?
About 1.6 GB at Q4_K_M with an 8K context window. That is the 1.0 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 Breeze-ASR-25 on an 8 GB GPU?
Yes, with F16: the 3.1 GB file needs about 4.0 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of Breeze-ASR-25 GGUF?
It depends on the quant: Q4_K_M is 1.0 GB, Q5_K_M is 1.2 GB, Q6_K is 1.3 GB, Q8_0 is 1.7 GB, BF16 is 3.1 GB, F16 is 3.1 GB.
Where can I download Breeze-ASR-25 GGUF?
The files are on Hugging Face in the handy-computer/Breeze-ASR-25-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 Breeze-ASR-25 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 Breeze-ASR-25 on 8 GB VRAM?
Yes, using Q4_K_M or Q5_K_M quantizations, which use about 1.6–1.8 GB of VRAM at 8 k context.
Does Breeze-ASR-25 support other languages?
The model retains Whisper’s 99‑language tokenizer, but only Taiwanese Mandarin and English are optimized and validated.