DeepSeek V2.5 needs about 142.4 GB of memory at Q3_K_S (a 101.7 GB file) with an 8K context window, which is more than a 48 GB GPU holds. This repository has 23 quants, from 47.4 GB (IQ1_S) to 250.6 GB (Q8_0). Longer context needs more: at 32K tokens the KV cache alone adds about 161.1 GB.
DeepSeek V2.5 – 230B parameter model with mixed‑expert architecture
DeepSeek V2.5 combines the general and coding abilities of DeepSeek‑V2‑Chat and DeepSeek‑Coder‑V2‑Instruct. The model is 230 billion parameters, trained on 163 840 context tokens. It is available in a variety of GGUF quantizations for llama.cpp inference.
Good at
- General conversation
- Coding assistance
- Instruction following
Limits
- Requires large VRAM for high‑quality quants
- No vision support
- License is other – check usage terms
Best for
- Chatbot development
- Code generation tools
- Instruction‑driven applications
Who it’s for. Developers and researchers who need a large, versatile language model for local deployment. The model is suitable for those who can manage the memory requirements of the chosen quantization.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| IQ1_S | 47.4 GB (2 parts) | 88.2 GB | Too big | Get file → | |
| IQ1_M | 52.7 GB (3 parts) | 93.4 GB | Too big | Get file → | |
| IQ2_XXS | 61.5 GB (3 parts) | 102.3 GB | Too big | Get file → | |
| IQ2_XS | 68.7 GB (3 parts) | 109.5 GB | Too big | Get file → | |
| IQ2_S | 69.9 GB (3 parts) | 110.6 GB | Too big | Get file → | |
| IQ2_M | 76.9 GB (4 parts) | 117.7 GB | Too big | Get file → | |
| Q2_K_S | 79.6 GB (4 parts) | 120.4 GB | Too big | Get file → | |
| Q2_K | 85.9 GB (4 parts) | 126.7 GB | Too big | Get file → | |
| IQ3_XXS | 90.8 GB (4 parts) | 131.6 GB | Too big | Get file → | |
| IQ3_XS | 96.3 GB (5 parts) | 137.1 GB | Too big | Get file → | |
| IQ3_S | 101.7 GB (5 parts) | 142.4 GB | Too big | Get file → | |
| Q3_K_S | 101.7 GB (5 parts) | 142.4 GB | Too big | Get file → | |
| IQ3_M | 103.4 GB (5 parts) | 144.1 GB | Too big | Get file → | |
| Q3_K | 112.7 GB (5 parts) | 153.4 GB | Too big | Get file → | |
| Q3_K_L | 122.4 GB (6 parts) | 163.1 GB | Too big | Get file → | |
| IQ4_XS | 125.6 GB (6 parts) | 166.3 GB | Too big | Get file → | |
| IQ4_NL | 132.9 GB (6 parts) | 173.7 GB | Too big | Get file → | |
| Q4_K_S | 133.9 GB (6 parts) | 174.6 GB | Too big | Get file → | |
| Q4_K | 142.5 GB (7 parts) | 183.2 GB | Too big | Get file → | |
| Q5_K_S | 162.3 GB (7 parts) | 203.1 GB | Too big | Get file → | |
| Q5_K | 167.2 GB (8 parts) | 208.0 GB | Too big | Get file → | |
| Q6_K | 193.5 GB (9 parts) | 234.3 GB | Too big | Get file → | |
| Q8_0 | 250.6 GB (11 parts) | 291.4 GB | Too big | Get file → |
For 8 GB VRAM, use Q2_K_S (≈120 GB memory at 8k context). For 16 GB VRAM, Q3_K_S (≈142 GB). For 32 GB VRAM, Q4_K (≈183 GB). Larger quants such as Q5_K or Q6_K require 200 GB+ VRAM and are not recommended for typical consumer GPUs.
What GPU do you need for DeepSeek V2.5?
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 DeepSeek V2.5 quant should you choose?
Q3_K_S (101.7 GB) is the middle option in this repository. A reasonable choice when a 4-bit file will not fit, mostly on larger models. Quality loss is noticeable but often acceptable.
If memory is tight, the smallest file is IQ1_S at 47.4 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 DeepSeek V2.5
How much VRAM does DeepSeek V2.5 need?
About 142.4 GB at Q3_K_S with an 8K context window. That is the 101.7 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 DeepSeek V2.5 on an 8 GB GPU?
Not at an 8K context. The smallest file needs about 88.2 GB, so you would need a larger GPU or CPU offloading.
What is the file size of DeepSeek V2.5 GGUF?
It depends on the quant: IQ1_S is 47.4 GB, IQ1_M is 52.7 GB, IQ2_XXS is 61.5 GB, IQ2_XS is 68.7 GB, IQ2_S is 69.9 GB, IQ2_M is 76.9 GB and 17 more.
What context length does DeepSeek V2.5 support?
The GGUF header lists a trained context length of 163,840 tokens (160K). The KV cache grows with the context you use: about 40.3 GB at 8K and 161.1 GB at 32K.
Where can I download DeepSeek V2.5 GGUF?
The files are on Hugging Face in the legraphista/DeepSeek-V2.5-IMat-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 DeepSeek V2.5 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 DeepSeek V2.5 on 8 GB VRAM?
Only the lowest‑quality quants such as IQ1_S or IQ1_M can fit, but they still need around 88–93 GB memory at 8k context, so 8 GB VRAM is insufficient.
Does the model support vision?
No, DeepSeek V2.5 does not have vision capabilities.
What license applies to the GGUF files?
The GGUF files are released under the MIT license, but the model license allows commercial use.