Kimi Linear 48B A3B Base needs about 39.4 GB of memory at Q4_K_M (a 29.7 GB file) with an 8K context window, so it fits on a 48 GB GPU. This repository has 24 quants, from 0.2 GB to 40.4 GB (Q6_K). Longer context needs more: at 32K tokens the KV cache alone adds about 36.7 GB.
Kimi Linear 48B A3B Base – a 49.1B parameter hybrid linear attention model
Kimi Linear 48B A3B Base is a 49.1B parameter model with 3B active parameters. It uses Kimi Delta Attention for efficient long‑context performance and supports up to 1,048,576 tokens of context. The model is released under the MIT license.
Good at
- Long‑context inference
- High decoding throughput
- Tool‑use support
Limits
- Requires large VRAM for full precision
- No vision capabilities
- Limited to text only
Best for
- Large‑scale text generation
- Long‑document summarization
- Reinforcement learning tasks
Who it’s for. Developers and researchers needing a high‑capacity model that excels at long‑context tasks. The model is suitable for those who can allocate sufficient GPU memory and prefer open‑source licensing.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| — | 0.2 GB | 9.8 GB | Too big | Get file → | |
| IQ1_S | 10.1 GB | 19.8 GB | Too big | Get file → | |
| IQ1_M | 11.2 GB | 20.9 GB | Too big | Get file → | |
| IQ2_XXS | 13.0 GB | 22.7 GB | Too big | Get file → | |
| IQ2_XS | 14.5 GB | 24.2 GB | Too big | Get file → | |
| IQ2_S | 14.7 GB | 24.3 GB | Too big | Get file → | |
| IQ2_M | 16.1 GB | 25.8 GB | Too big | Get file → | |
| Q2_K_S | 16.7 GB | 26.3 GB | Too big | Get file → | |
| Q2_K | 18.0 GB | 27.7 GB | Too big | Get file → | |
| IQ3_XXS | 19.0 GB | 28.7 GB | Too big | Get file → | |
| IQ3_XS | 20.2 GB | 29.8 GB | Too big | Get file → | |
| IQ3_S | 21.3 GB | 31.0 GB | Too big | Get file → | |
| Q3_K_S | 21.3 GB | 31.0 GB | Too big | Get file → | |
| IQ3_M | 21.5 GB | 31.2 GB | Too big | Get file → | |
| Q3_K_M | 23.5 GB | 33.2 GB | Too big | Get file → | |
| Q3_K_L | 25.5 GB | 35.2 GB | Too big | Get file → | |
| IQ4_XS | 26.3 GB | 35.9 GB | Too big | Get file → | |
| Q4_0 | 27.9 GB | 37.5 GB | Too big | Get file → | |
| Q4_K_S | 28.0 GB | 37.6 GB | Too big | Get file → | |
| Q4_K_M | 29.7 GB | 39.4 GB | Too big | Get file → | |
| Q4_1 | 30.8 GB | 40.5 GB | Too big | Get file → | |
| Q5_K_S | 33.9 GB | 43.6 GB | Too big | Get file → | |
| Q5_K_M | 34.9 GB | 44.5 GB | Too big | Get file → | |
| Q6_K | 40.4 GB | 50.0 GB | Too big | Get file → |
For 8 GB VRAM, use IQ1_S (file size 10.08 GB, memory 19.8 GB at 8 k context) or IQ1_M (11.19 GB, 20.9 GB). For 16 GB VRAM, IQ2_XS (14.52 GB, 24.2 GB) or Q2_K (18.03 GB, 27.7 GB) are suitable. For 32 GB VRAM, Q4_K_M (29.7 GB, 39.4 GB) or Q5_K_M (34.87 GB, 44.5 GB) provide higher quality.
What GPU do you need for Kimi Linear 48B A3B Base?
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: no quant fits at an 8K context.
- 6 GB VRAM: no quant fits at an 8K context.
- 8 GB VRAM: no quant fits at an 8K context.
- 12 GB VRAM: the best quant that fits is the main file (0.2 GB file, about 9.8 GB in use).
- 16 GB VRAM: the best quant that fits is the main file (0.2 GB file, about 9.8 GB in use).
- 24 GB VRAM: the best quant that fits is IQ2_XXS (13.0 GB file, about 22.7 GB in use).
- 32 GB VRAM: the best quant that fits is IQ3_M (21.5 GB file, about 31.2 GB in use).
- 48 GB VRAM: the best quant that fits is Q5_K_M (34.9 GB file, about 44.5 GB in use).
Which Kimi Linear 48B A3B Base quant should you choose?
Q4_K_M is the usual starting point for Kimi Linear 48B A3B Base: it balances file size and quality well. Here it is a 29.7 GB file.
If memory is tight, the smallest file is the first one in the table at 0.2 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 Kimi Linear 48B A3B Base
How much VRAM does Kimi Linear 48B A3B Base need?
About 39.4 GB at Q4_K_M with an 8K context window. That is the 29.7 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 Kimi Linear 48B A3B Base on an 8 GB GPU?
Not at an 8K context. The smallest file needs about 9.8 GB, so you would need a larger GPU or CPU offloading.
What is the file size of Kimi Linear 48B A3B Base GGUF?
It depends on the quant: file is 0.2 GB, IQ1_S is 10.1 GB, IQ1_M is 11.2 GB, IQ2_XXS is 13.0 GB, IQ2_XS is 14.5 GB, IQ2_S is 14.7 GB and 18 more.
What context length does Kimi Linear 48B A3B Base support?
The GGUF header lists a trained context length of 1,048,576 tokens (1024K). The KV cache grows with the context you use: about 9.2 GB at 8K and 36.7 GB at 32K.
Where can I download Kimi Linear 48B A3B Base GGUF?
The files are on Hugging Face in the mradermacher/Kimi-Linear-48B-A3B-Base-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 Kimi Linear 48B A3B Base 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 Kimi Linear 48B on 8 GB VRAM?
No, the smallest quant requires at least 19.8 GB VRAM at 8 k context. 8 GB VRAM is insufficient.
Does the model support image inputs?
No, the model has no vision capabilities.
What license does the model use?
The model is released under the MIT license.