LFM2.5 Audio 1.5B JP needs about 0.9 GB of memory at F32 (a 0.3 GB file) with an 8K context window, so it fits on a 4 GB GPU. This repository has 12 quants, from 0.0 GB (Q4_0) to 4.7 GB (F32). Longer context needs more: at 32K tokens the KV cache alone adds about 0.5 GB.
LFM2.5 Audio 1.5B JP – lightweight Japanese multimodal speech model
LFM2.5 Audio 1.5B JP is a 1.5 billion‑parameter model that handles Japanese speech and text. It supports real‑time speech‑to‑speech and non‑conversational ASR/TTS tasks. The model is built on LFM2.5 with a FastConformer audio encoder and RQ‑transformer.
Good at
- Real‑time Japanese speech‑to‑speech
- Low‑latency conversational audio
- Multimodal audio‑text generation
Limits
- No vision support
- Requires 8 k context for estimates
- License is other
Best for
- Japanese voice assistants
- Audio transcription in Japanese
- Real‑time Japanese chatbots
Who it’s for. Developers building Japanese audio applications who need a small, low‑latency model. The model is suitable for edge devices and real‑time interaction scenarios.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| Q4_0 | 0.0 GB | 0.7 GB | Fits | Get file → | |
| Q8_0 | 0.1 GB | 0.7 GB | Fits | Get file → | |
| Q4_0 | 0.1 GB | 0.7 GB | Fits | Get file → | |
| F16 | 0.1 GB | 0.8 GB | Fits | Get file → | |
| Q8_0 | 0.2 GB | 0.8 GB | Fits | Get file → | |
| F32 | 0.3 GB | 0.9 GB | Fits | Get file → | |
| F16 | 0.4 GB | 1.0 GB | Fits | Get file → | |
| Q4_0 | 0.7 GB | 1.3 GB | Fits | Get file → | |
| F32 | 0.8 GB | 1.4 GB | Fits | Get file → | |
| Q8_0 | 1.2 GB | 1.9 GB | Fits | Get file → | |
| F16 | 2.3 GB | 3.0 GB | Fits | Get file → | |
| F32 | 4.7 GB | 5.3 GB | Fits | Get file → |
For 8 GB VRAM, use Q4_0 (0.05 GB) or Q8_0 (0.07 GB). For 16 GB VRAM, Q4_0 (0.11 GB) or F16 (0.14 GB) are suitable. For 32 GB VRAM, F32 (0.28 GB) or larger quantizations can be used. Memory estimates are approximate.
What GPU do you need for LFM2.5 Audio 1.5B JP?
These figures add the file size, an fp16 KV cache at an 8K context window (worked out from this model’s own header) 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 (2.3 GB file, about 3.0 GB in use).
- 6 GB VRAM: the best quant that fits is F32 (4.7 GB file, about 5.3 GB in use).
Which LFM2.5 Audio 1.5B JP quant should you choose?
F32 (0.3 GB) is the middle option in this repository. The unquantized model. It gives original quality and needs the most memory, so it is mainly for conversion or reference.
If memory is tight, the smallest file is Q4_0 at 0.0 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 LFM2.5 Audio 1.5B JP
How much VRAM does LFM2.5 Audio 1.5B JP need?
About 0.9 GB at F32 with an 8K context window. That is the 0.3 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 LFM2.5 Audio 1.5B JP on an 8 GB GPU?
Yes, with F32: the 4.7 GB file needs about 5.3 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of LFM2.5 Audio 1.5B JP GGUF?
It depends on the quant: Q4_0 is 0.0 GB, Q8_0 is 0.1 GB, Q4_0 is 0.1 GB, F16 is 0.1 GB, Q8_0 is 0.2 GB, F32 is 0.3 GB and 6 more.
What context length does LFM2.5 Audio 1.5B JP support?
The GGUF header lists a trained context length of 128,000 tokens (125K). The KV cache grows with the context you use: about 0.1 GB at 8K and 0.5 GB at 32K.
Where can I download LFM2.5 Audio 1.5B JP GGUF?
The files are on Hugging Face in the LiquidAI/LFM2.5-Audio-1.5B-JP-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 LFM2.5 Audio 1.5B JP use?
Hugging Face lists the license as other. 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 LFM2.5 Audio 1.5B JP on 8 GB VRAM?
Yes, the Q4_0 and Q8_0 quantizations are 0.05 GB and 0.07 GB respectively, well within 8 GB VRAM.
Does this model support vision tasks?
No, the model only detects vision ability but does not provide vision inference.
What context length does it support?
It was trained with 128000 tokens and estimates are given for 8 k context.