DE

DeepSeek V2.5

DeepSeekGGUF by second-stateUpdated 2024-10-16
Download on Hugging Face
Parameters236B
Context160K
Layers60
Architecturedeepseek2
Smallest quant85.9 GB
Quick answer

DeepSeek V2.5 needs about 183.2 GB of memory at Q4_K_M (a 142.5 GB file) with an 8K context window, which is more than a 48 GB GPU holds. This repository has 11 quants, from 85.9 GB (Q2_K) to 471.6 GB (F16). Longer context needs more: at 32K tokens the KV cache alone adds about 161.1 GB.

DeepSeek V2.5 is a 236B-parameter DeepSeek model in GGUF format, quantized by second-state. Its trained context length is 160K tokens. The repository has 11 quant files, from 85.9 GB (Q2_K) to 471.6 GB (F16).

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
Q2_K85.9 GB (3 parts) 126.7 GB
Too big Get file →
Q3_K_S101.7 GB (4 parts) 142.4 GB
Too big Get file →
Q3_K_M112.7 GB (4 parts) 153.4 GB
Too big Get file →
Q3_K_L122.4 GB (5 parts) 163.1 GB
Too big Get file →
Q4_0132.9 GB (5 parts) 173.7 GB
Too big Get file →
Q4_K_S133.9 GB (5 parts) 174.6 GB
Too big Get file →
Q4_K_M142.5 GB (5 parts) 183.2 GB
Too big Get file →
Q5_0162.3 GB (6 parts) 203.1 GB
Too big Get file →
Q5_K_S162.3 GB (6 parts) 203.1 GB
Too big Get file →
Q5_K_M167.2 GB (6 parts) 208.0 GB
Too big Get file →
F16471.6 GB (17 parts) 512.4 GB
Too big Get file →

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?

Q4_K_M is the usual starting point for DeepSeek V2.5: it balances file size and quality well. Here it is a 142.5 GB file.

If memory is tight, the smallest file is Q2_K at 85.9 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

Architecturedeepseek2
Parameters236B (approximate, mixture of experts)
Layers60
Trained context163,840 tokens
KV cache (fp16)4800 KB per token

Source

Quantized bysecond-state
FamilyDeepSeek
Licenseother
Downloads (HF)393
Last updated2024-10-16

Frequently asked questions about DeepSeek V2.5

How much VRAM does DeepSeek V2.5 need?

About 183.2 GB at Q4_K_M with an 8K context window. That is the 142.5 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 126.7 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: Q2_K is 85.9 GB, Q3_K_S is 101.7 GB, Q3_K_M is 112.7 GB, Q3_K_L is 122.4 GB, Q4_0 is 132.9 GB, Q4_K_S is 133.9 GB and 5 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 second-state/DeepSeek-V2.5-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.

Other GGUFs of DeepSeek-V2.5

Similar models

Browse all →

More from second-state

All by second-state →

More DeepSeek models

All DeepSeek →

Other large models

All large models →