Kisoku 3b Sft needs about 3.4 GB of memory at Q4_K_M (a 2.2 GB file) with an 8K context window, so it fits on a 4 GB GPU. This repository has 25 quants, from 0.0 GB to 2.9 GB (Q6_K).
Kisoku 3b Sft – a 3.56B parameter Granite model fine‑tuned for instruction following
Kisoku 3b Sft is a 3.56B parameter model built on the Granite architecture. It was supervised fine‑tuned from the Kisoku 3B base checkpoint using a custom text‑only chat template. The model is available in GGUF format for local deployment via llama.cpp.
Good at
- Instruction following
- Chat‑style interactions
- Text generation
Limits
- Requires temperature adjustment due to Granite logit scaling
- No vision or RLHF safety alignment
- English‑focused
Best for
- Local chatbots on modest GPUs
- Rapid prototyping of instruction‑tuned LLMs
- Educational demonstrations of Granite architecture
Who it’s for. Developers and researchers who need a lightweight instruction‑tuned model that can run locally with llama.cpp. The model is suitable for those who prefer open‑source licensing and are comfortable adjusting inference temperature.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| — | 0.0 GB | 0.9 GB | Fits | Get file → | |
| IQ1_S | 1.0 GB | 1.9 GB | Fits | Get file → | |
| IQ1_M | 1.1 GB | 1.9 GB | Fits | Get file → | |
| IQ2_XXS | 1.1 GB | 2.0 GB | Fits | Get file → | |
| IQ2_XS | 1.2 GB | 2.1 GB | Fits | Get file → | |
| IQ2_S | 1.3 GB | 2.2 GB | Fits | Get file → | |
| IQ2_M | 1.4 GB | 2.2 GB | Fits | Get file → | |
| Q2_K_S | 1.4 GB | 2.3 GB | Fits | Get file → | |
| Q2_K | 1.5 GB | 2.3 GB | Fits | Get file → | |
| IQ3_XXS | 1.5 GB | 2.4 GB | Fits | Get file → | |
| IQ3_XS | 1.6 GB | 2.5 GB | Fits | Get file → | |
| Q3_K_S | 1.7 GB | 2.5 GB | Fits | Get file → | |
| IQ3_S | 1.7 GB | 2.6 GB | Fits | Get file → | |
| IQ3_M | 1.7 GB | 2.6 GB | Fits | Get file → | |
| Q3_K_M | 1.8 GB | 2.7 GB | Fits | Get file → | |
| Q3_K_L | 2.0 GB | 2.8 GB | Fits | Get file → | |
| IQ4_XS | 2.0 GB | 2.9 GB | Fits | Get file → | |
| Q4_0 | 2.1 GB | 3.0 GB | Fits | Get file → | |
| IQ4_NL | 2.1 GB | 3.0 GB | Fits | Get file → | |
| Q4_K_S | 2.1 GB | 3.0 GB | Fits | Get file → | |
| Q4_K_M | 2.2 GB | 3.1 GB | Fits | Get file → | |
| Q4_1 | 2.3 GB | 3.2 GB | Fits | Get file → | |
| Q5_K_S | 2.5 GB | 3.4 GB | Fits | Get file → | |
| Q5_K_M | 2.6 GB | 3.4 GB | Fits | Get file → | |
| Q6_K | 2.9 GB | 3.8 GB | Fits | Get file → |
For 8 GB VRAM, use Q4_0 (file size 2.12 GB, estimated 3.3 GB memory at 8 k context). If 12 GB VRAM is available, Q5_K_M (2.56 GB, 3.8 GB memory) offers higher quality. For 16 GB VRAM, Q6_K (2.93 GB, 4.1 GB memory) provides the best trade‑off between size and performance.
What GPU do you need for Kisoku 3b Sft?
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 Q5_K_M (2.6 GB file, about 3.8 GB in use).
- 6 GB VRAM: the best quant that fits is Q6_K (2.9 GB file, about 4.1 GB in use).
Which Kisoku 3b Sft quant should you choose?
Q4_K_M is the usual starting point for Kisoku 3b Sft: it balances file size and quality well. Here it is a 2.2 GB file.
If memory is tight, the smallest file is the first one in the table at 0.0 GB, with the quality trade-offs that come with fewer bits.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about Kisoku 3b Sft
How much VRAM does Kisoku 3b Sft need?
About 3.4 GB at Q4_K_M with an 8K context window. That is the 2.2 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 Kisoku 3b Sft on an 8 GB GPU?
Yes, with Q6_K: the 2.9 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 Kisoku 3b Sft GGUF?
It depends on the quant: file is 0.0 GB, IQ1_S is 1.0 GB, IQ1_M is 1.1 GB, IQ2_XXS is 1.1 GB, IQ2_XS is 1.2 GB, IQ2_S is 1.3 GB and 19 more.
What context length does Kisoku 3b Sft support?
The GGUF header lists a trained context length of 4,096 tokens (4K). The KV cache grows with the context you use: about 0.7 GB at 8K and 0.4 GB at 4K.
Where can I download Kisoku 3b Sft GGUF?
The files are on Hugging Face in the mradermacher/kisoku-3b-sft-i1-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 Kisoku 3b Sft 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 Kisoku 3b Sft on 8 GB VRAM?
Yes, the Q4_0 quant requires about 3.3 GB of VRAM at 8 k context, which fits within an 8 GB budget.
Does this model support vision tasks?
No, the model is text‑only and does not include vision capabilities.
What temperature should I use for inference?
Because of Granite logit scaling, a temperature around 0.01 is recommended for standard behavior.