Kimi-K2-Instruct-0905 needs about 622.8 GB of memory at Q4_K_M (a 621.2 GB file) with an 8K context window, which is more than a 48 GB GPU holds. This repository has 27 quants, from 248.6 GB (TQ1_0) to 2053.2 GB (BF16). Longer context needs more: at 32K tokens the KV cache alone adds about 4.3 GB.
Kimi-K2-Instruct-0905: 1 trillion‑parameter MoE model with 256k‑token context
Kimi-K2-Instruct-0905 is a 1 trillion‑parameter mixture‑of‑experts model built on deepseek2 architecture. It offers a 256k‑token context window and 32 billion activated parameters. The model is available in multiple GGUF quantizations.
Good at
- Long‑form coding assistance
- Large context reasoning
- Agentic task execution
Limits
- Requires high VRAM for large quants
- No vision support
- License type is other
Best for
- Complex software development
- Extended document summarization
- Multi‑step reasoning tasks
Who it’s for. Developers and researchers needing a large‑context, MoE language model for coding and reasoning. The model is suited for environments that can support its memory demands.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| TQ1_0 | 248.6 GB (6 parts) | 250.2 GB | Too big | Get file → | |
| IQ1_S | 281.9 GB (6 parts) | 283.5 GB | Too big | Get file → | |
| IQ1_M | 304.8 GB (7 parts) | 306.4 GB | Too big | Get file → | |
| IQ2_XXS | 330.8 GB (7 parts) | 332.4 GB | Too big | Get file → | |
| IQ2_M | 349.1 GB (8 parts) | 350.7 GB | Too big | Get file → | |
| Q2_K | 374.1 GB (8 parts) | 375.7 GB | Too big | Get file → | |
| Q2_K_L | 374.4 GB (8 parts) | 376.0 GB | Too big | Get file → | |
| Q2_K_XL | 383.6 GB (8 parts) | 385.2 GB | Too big | Get file → | |
| IQ3_XXS | 417.3 GB (9 parts) | 418.9 GB | Too big | Get file → | |
| Q3_K_S | 443.2 GB (10 parts) | 444.7 GB | Too big | Get file → | |
| Q3_K_XL | 452.3 GB (10 parts) | 453.9 GB | Too big | Get file → | |
| Q3_K_M | 490.0 GB (11 parts) | 491.6 GB | Too big | Get file → | |
| IQ4_XS | 547.2 GB (12 parts) | 548.8 GB | Too big | Get file → | |
| IQ4_NL | 579.1 GB (12 parts) | 580.7 GB | Too big | Get file → | |
| Q4_0 | 581.2 GB (13 parts) | 582.8 GB | Too big | Get file → | |
| Q4_K_S | 583.3 GB (13 parts) | 584.9 GB | Too big | Get file → | |
| Q4_K_XL | 588.1 GB (13 parts) | 589.7 GB | Too big | Get file → | |
| Q4_K_M | 621.2 GB (13 parts) | 622.8 GB | Too big | Get file → | |
| Q4_1 | 642.9 GB (13 parts) | 644.5 GB | Too big | Get file → | |
| Q5_K_S | 707.0 GB (15 parts) | 708.6 GB | Too big | Get file → | |
| Q5_K_M | 728.7 GB (16 parts) | 730.3 GB | Too big | Get file → | |
| Q5_K_XL | 734.1 GB (16 parts) | 735.6 GB | Too big | Get file → | |
| Q6_K | 842.9 GB (18 parts) | 844.5 GB | Too big | Get file → | |
| Q6_K_XL | 879.1 GB (19 parts) | 880.6 GB | Too big | Get file → | |
| Q8_0 | 1091.1 GB (23 parts) | 1092.6 GB | Too big | Get file → | |
| Q8_K_XL | 1190.1 GB (25 parts) | 1191.6 GB | Too big | Get file → | |
| BF16 | 2053.2 GB (46 parts) | 2054.7 GB | Too big | Get file → |
For 8 GB VRAM, use Q4_K_S or Q4_K_XL; memory estimates are around 584.9 GB at 8k context. For 16 GB VRAM, Q4_K_M or Q4_1 are suitable. For larger budgets, Q5_K_S or Q5_K_M provide higher quality.
What GPU do you need for Kimi-K2-Instruct-0905?
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: no quant fits at an 8K context.
- 16 GB VRAM: no quant fits at an 8K context.
- 24 GB VRAM: no quant fits at an 8K context.
- 32 GB VRAM: no quant fits at an 8K context.
- 48 GB VRAM: no quant fits at an 8K context.
Which Kimi-K2-Instruct-0905 quant should you choose?
Q4_K_M is the usual starting point for Kimi-K2-Instruct-0905: it balances file size and quality well. Here it is a 621.2 GB file.
The Q8_0 file is 1091.1 GB, about 1.8 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 TQ1_0 at 248.6 GB, with the quality trade-offs that come with fewer bits. Lower bit counts mean smaller files and more quality loss.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about Kimi-K2-Instruct-0905
How much VRAM does Kimi-K2-Instruct-0905 need?
About 622.8 GB at Q4_K_M with an 8K context window. That is the 621.2 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 Kimi-K2-Instruct-0905 on an 8 GB GPU?
Not at an 8K context. The smallest file needs about 250.2 GB, so you would need a larger GPU or CPU offloading.
What is the file size of Kimi-K2-Instruct-0905 GGUF?
It depends on the quant: TQ1_0 is 248.6 GB, IQ1_S is 281.9 GB, IQ1_M is 304.8 GB, IQ2_XXS is 330.8 GB, IQ2_M is 349.1 GB, Q2_K is 374.1 GB and 21 more.
What context length does Kimi-K2-Instruct-0905 support?
The GGUF header lists a trained context length of 262,144 tokens (256K). The KV cache grows with the context you use: about 1.1 GB at 8K and 4.3 GB at 32K.
Where can I download Kimi-K2-Instruct-0905 GGUF?
The files are on Hugging Face in the unsloth/Kimi-K2-Instruct-0905-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-K2-Instruct-0905 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 Kimi-K2-Instruct-0905 on 8 GB VRAM?
Only the smallest quants such as Q4_K_S or Q4_K_XL can fit, but performance will be limited.
Does the model support vision tasks?
No, Kimi-K2-Instruct-0905 does not have vision capabilities.
What context length does the model support?
The model supports a context window of 256k tokens.