mochi needs about 6.8 GB of memory at Q4_K_M (a 5.6 GB file) with an 8K context window, so it fits on a 8 GB GPU. This repository has 13 quants, from 0.9 GB (F16) to 8.2 GB (Q6_K).
mochi is a 10B-parameter Mochi model in GGUF format, quantized by calcuis. The repository has 13 quant files, from 0.9 GB (F16) to 8.2 GB (Q6_K).
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| F16 | 0.9 GB | 1.5 GB | Fits | Get file → | |
| Q4_0 | 2.8 GB | 3.6 GB | Fits | Get file → | |
| Q2_K | 3.3 GB | 4.2 GB | Fits | Get file → | |
| Q3_K_M | 4.3 GB | 5.3 GB | Fits | Get file → | |
| IQ3_XXS | 4.7 GB | 5.7 GB | Fits | Get file → | |
| Q8_0 | 5.2 GB | 6.3 GB | Fits | Get file → | |
| Q4_K_M | 5.6 GB | 6.8 GB | Fits | Get file → | |
| IQ4_XS | 5.7 GB | 6.9 GB | Fits | Get file → | |
| IQ4_NL | 6.0 GB | 7.2 GB | Tight | Get file → | |
| Q4_0 | 6.0 GB | 7.2 GB | Tight | Get file → | |
| Q5_K_M | 6.9 GB | 8.2 GB | Too big | Get file → | |
| Q5_0 | 7.2 GB | 8.6 GB | Too big | Get file → | |
| Q6_K | 8.2 GB | 9.7 GB | Too big | Get file → |
What GPU do you need for mochi?
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_0 (2.8 GB file, about 3.6 GB in use).
- 6 GB VRAM: the best quant that fits is IQ3_XXS (4.7 GB file, about 5.7 GB in use).
- 8 GB VRAM: the best quant that fits is IQ4_NL (6.0 GB file, about 7.2 GB in use).
- 12 GB VRAM: the best quant that fits is Q6_K (8.2 GB file, about 9.7 GB in use).
Which mochi quant should you choose?
Q4_K_M is the usual starting point for mochi: it balances file size and quality well. Here it is a 5.6 GB file.
The Q8_0 file is 5.2 GB, about 0.9 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.
If memory is tight, the smallest file is F16 at 0.9 GB, with the quality trade-offs that come with fewer bits. The unquantized model. It gives original quality and needs the most memory, so it is mainly for conversion or reference.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about mochi
How much VRAM does mochi need?
About 6.8 GB at Q4_K_M with an 8K context window. That is the 5.6 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 mochi on an 8 GB GPU?
Yes, with IQ4_NL: the 6.0 GB file needs about 7.2 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of mochi GGUF?
It depends on the quant: F16 is 0.9 GB, Q4_0 is 2.8 GB, Q2_K is 3.3 GB, Q3_K_M is 4.3 GB, IQ3_XXS is 4.7 GB, Q8_0 is 5.2 GB and 7 more.
Where can I download mochi GGUF?
The files are on Hugging Face in the calcuis/mochi-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 mochi 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.