Sarashina2.2 3b Instruct v0.1 needs about 3.9 GB of memory at Q4_K_M (a 2.1 GB file) with an 8K context window, so it fits on a 4 GB GPU. This repository has 24 quants, from 0.9 GB (IQ1_S) to 3.6 GB (Q8_0).
Sarashina2.2 3b Instruct v0.1 – a 3.36B Japanese instruction model
Sarashina2.2 3b Instruct v0.1 is a 3.36B parameter autoregressive language model focused on Japanese instruction following. It was trained with 8192 context tokens and released under the MIT license. The model is available in multiple GGUF quantizations.
Good at
- Japanese instruction following
- Short to medium context generation
- Low‑memory inference
Limits
- No vision support
- Limited safety training
- Requires 8k context for full capability
Best for
- Chatbots in Japanese
- Content drafting in Japanese
- Japanese code generation
Who it’s for. Developers building Japanese language applications who need a lightweight instruction model. The model is suitable for users who can fine‑tune or prompt for safety and privacy concerns.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| IQ1_S | 0.9 GB | 2.7 GB | Fits | Get file → | |
| IQ1_M | 0.9 GB | 2.8 GB | Fits | Get file → | |
| IQ2_XXS | 1.0 GB | 2.9 GB | Fits | Get file → | |
| IQ2_XS | 1.1 GB | 2.9 GB | Fits | Get file → | |
| IQ2_S | 1.2 GB | 3.0 GB | Fits | Get file → | |
| IQ2_M | 1.2 GB | 3.1 GB | Fits | Get file → | |
| Q2_K | 1.3 GB | 3.2 GB | Fits | Get file → | |
| IQ3_XXS | 1.4 GB | 3.2 GB | Fits | Get file → | |
| IQ3_XS | 1.5 GB | 3.3 GB | Fits | Get file → | |
| Q3_K_S | 1.5 GB | 3.4 GB | Fits | Get file → | |
| IQ3_S | 1.5 GB | 3.4 GB | Fits | Get file → | |
| IQ3_M | 1.6 GB | 3.4 GB | Fits | Get file → | |
| Q3_K_M | 1.7 GB | 3.5 GB | Fits | Get file → | |
| Q3_K_L | 1.8 GB | 3.7 GB | Fits | Get file → | |
| IQ4_XS | 1.9 GB | 3.7 GB | Fits | Get file → | |
| Q4_0 | 2.0 GB | 3.8 GB | Fits | Get file → | |
| IQ4_NL | 2.0 GB | 3.8 GB | Fits | Get file → | |
| Q4_K_S | 2.0 GB | 3.8 GB | Fits | Get file → | |
| Q4_K_M | 2.1 GB | 3.9 GB | Fits | Get file → | |
| Q5_0 | 2.3 GB | 4.2 GB | Fits | Get file → | |
| Q5_K_S | 2.3 GB | 4.2 GB | Fits | Get file → | |
| Q5_K_M | 2.4 GB | 4.2 GB | Fits | Get file → | |
| Q6_K | 2.8 GB | 4.6 GB | Fits | Get file → | |
| Q8_0 | 3.6 GB | 5.4 GB | Fits | Get file → |
For 8 GB VRAM, use IQ1_S (0.86 GB) or IQ1_M (0.92 GB). For 12 GB VRAM, IQ2_M (1.25 GB) or Q3_K_M (1.69 GB) are good choices. Higher precision quants like Q5_K_M (2.4 GB) or Q6_K (2.76 GB) require 16 GB or more. Estimates are approximate.
What GPU do you need for Sarashina2.2 3b Instruct v0.1?
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 Q4_K_M (2.1 GB file, about 3.9 GB in use).
- 6 GB VRAM: the best quant that fits is Q8_0 (3.6 GB file, about 5.4 GB in use).
Which Sarashina2.2 3b Instruct v0.1 quant should you choose?
Q4_K_M is the usual starting point for Sarashina2.2 3b Instruct v0.1: it balances file size and quality well. Here it is a 2.1 GB file.
The Q8_0 file is 3.6 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.
If memory is tight, the smallest file is IQ1_S at 0.9 GB, with the quality trade-offs that come with fewer bits. Use it only when memory is very tight. Expect a clear drop in quality, which is usually worse on small models and easier to live with on large ones.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about Sarashina2.2 3b Instruct v0.1
How much VRAM does Sarashina2.2 3b Instruct v0.1 need?
About 3.9 GB at Q4_K_M with an 8K context window. That is the 2.1 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 Sarashina2.2 3b Instruct v0.1 on an 8 GB GPU?
Yes, with Q8_0: the 3.6 GB file needs about 5.4 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of Sarashina2.2 3b Instruct v0.1 GGUF?
It depends on the quant: IQ1_S is 0.9 GB, IQ1_M is 0.9 GB, IQ2_XXS is 1.0 GB, IQ2_XS is 1.1 GB, IQ2_S is 1.2 GB, IQ2_M is 1.2 GB and 18 more.
What context length does Sarashina2.2 3b Instruct v0.1 support?
The GGUF header lists a trained context length of 8,192 tokens (8K). The KV cache grows with the context you use: about 1.3 GB at 8K and 1.3 GB at 8K.
Where can I download Sarashina2.2 3b Instruct v0.1 GGUF?
The files are on Hugging Face in the mmnga/sarashina2.2-3b-instruct-v0.1-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 Sarashina2.2 3b Instruct v0.1 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 this model on 8 GB VRAM?
Yes, the IQ1_S and IQ1_M quantizations fit within 8 GB VRAM with an estimated 2.7–2.8 GB memory usage at 8k context.
Does the model support English?
The model is trained primarily for Japanese; English performance is limited and not guaranteed.
Is the model safe for production use?
The model has limited safety training and may produce inaccurate or biased outputs. It is recommended to fine‑tune or filter outputs for production.