DE

Deepseek-R1-0528

DeepSeekGGUF by unslothUpdated 2025-06-15
Download on Hugging Face
Parameters670B
Context160K
Layers61
Architecturedeepseek2
Smallest quant161.6 GB
Quick answer

Deepseek-R1-0528 needs about 406.5 GB of memory at Q4_K_M (a 404.9 GB file) with an 8K context window, which is more than a 48 GB GPU holds. This repository has 27 quants, from 161.6 GB (TQ1_0) to 1342.3 GB (BF16). Longer context needs more: at 32K tokens the KV cache alone adds about 4.3 GB.

DeepSeek-R1-0528: 670 Billion‑parameter reasoning model in GGUF format

DeepSeek-R1-0528 is a 670 Billion‑parameter model optimized for reasoning and coding. It supports a 163 840‑token context window and is released under the MIT license. The model is available in multiple GGUF quantizations for different memory budgets.

Good at

  • Advanced reasoning tasks
  • Complex coding problems
  • Large‑context text generation

Limits

  • Requires 8‑bit or higher quant for full context
  • No vision or audio capabilities
  • Large file sizes for high‑precision quants

Best for

  • Academic research on reasoning
  • Developing function‑calling agents
  • Large‑context document summarization

Who it’s for. Researchers and developers needing a high‑capacity reasoning model that can be run locally with llama.cpp. The model is suitable for users who can manage large file sizes and have sufficient GPU or CPU memory.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
TQ1_0161.6 GB 163.2 GB
Too big Get file →
IQ1_S185.5 GB (4 parts) 187.1 GB
Too big Get file →
IQ1_M200.5 GB (5 parts) 202.0 GB
Too big Get file →
IQ2_XXS216.8 GB (5 parts) 218.4 GB
Too big Get file →
IQ2_M228.5 GB (5 parts) 230.1 GB
Too big Get file →
Q2_K244.8 GB (5 parts) 246.4 GB
Too big Get file →
Q2_K_L245.0 GB (5 parts) 246.6 GB
Too big Get file →
Q2_K_XL251.1 GB (6 parts) 252.7 GB
Too big Get file →
IQ3_XXS272.9 GB (6 parts) 274.5 GB
Too big Get file →
Q3_K_S289.7 GB (6 parts) 291.3 GB
Too big Get file →
Q3_K_XL295.9 GB (7 parts) 297.5 GB
Too big Get file →
Q3_K_M319.8 GB (7 parts) 321.4 GB
Too big Get file →
IQ4_XS357.7 GB (8 parts) 359.3 GB
Too big Get file →
IQ4_NL378.6 GB (8 parts) 380.2 GB
Too big Get file →
Q4_0379.5 GB (8 parts) 381.1 GB
Too big Get file →
Q4_K_S380.5 GB (8 parts) 382.1 GB
Too big Get file →
Q4_K_XL384.0 GB (8 parts) 385.6 GB
Too big Get file →
Q4_K_M404.9 GB (9 parts) 406.5 GB
Too big Get file →
Q4_1420.4 GB (9 parts) 422.0 GB
Too big Get file →
Q5_K_S462.2 GB (10 parts) 463.8 GB
Too big Get file →
Q5_K_M475.8 GB (10 parts) 477.4 GB
Too big Get file →
Q5_K_XL481.4 GB (10 parts) 482.9 GB
Too big Get file →
Q6_K551.0 GB (12 parts) 552.6 GB
Too big Get file →
Q6_K_XL574.9 GB (12 parts) 576.5 GB
Too big Get file →
Q8_0713.3 GB (15 parts) 714.9 GB
Too big Get file →
Q8_K_XL780.6 GB (16 parts) 782.2 GB
Too big Get file →
BF161342.3 GB (30 parts) 1343.9 GB
Too big Get file →

For 8 GB VRAM, use Q4_K_S (≈382 GB file, 382 GB memory at 8k context). For 16 GB VRAM, Q4_K_M (≈406 GB memory) or Q5_K_S (≈464 GB memory) are suitable. For 32 GB VRAM, Q5_K_M or Q5_K_XL provide higher precision with acceptable memory usage.

What GPU do you need for Deepseek-R1-0528?

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-R1-0528 quant should you choose?

Q4_K_M is the usual starting point for Deepseek-R1-0528: it balances file size and quality well. Here it is a 404.9 GB file.

The Q8_0 file is 713.3 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 161.6 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

Architecturedeepseek2
Parameters670B (approximate, mixture of experts)
Layers61
Trained context163,840 tokens
KV cache (fp16)130 KB per token

Source

Quantized byunsloth
FamilyDeepSeek
Licensemit
Downloads (HF)14,608
Last updated2025-06-15

Frequently asked questions about Deepseek-R1-0528

How much VRAM does Deepseek-R1-0528 need?

About 406.5 GB at Q4_K_M with an 8K context window. That is the 404.9 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-R1-0528 on an 8 GB GPU?

Not at an 8K context. The smallest file needs about 163.2 GB, so you would need a larger GPU or CPU offloading.

What is the file size of Deepseek-R1-0528 GGUF?

It depends on the quant: TQ1_0 is 161.6 GB, IQ1_S is 185.5 GB, IQ1_M is 200.5 GB, IQ2_XXS is 216.8 GB, IQ2_M is 228.5 GB, Q2_K is 244.8 GB and 21 more.

What context length does Deepseek-R1-0528 support?

The GGUF header lists a trained context length of 163,840 tokens (160K). 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 Deepseek-R1-0528 GGUF?

The files are on Hugging Face in the unsloth/DeepSeek-R1-0528-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-R1-0528 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 DeepSeek-R1-0528 on 8 GB VRAM?

Use the Q4_K_S quant, which requires about 382 GB of VRAM at 8k context. It is the smallest quant that supports the full context length.

Does the model support vision or audio input?

No. The model is text‑only and does not include vision or audio capabilities.

What license does the model use?

The model is released under the MIT license, allowing free use and modification.

Other GGUFs of DeepSeek-R1-0528

Similar models

Browse all →

More from unsloth

All by unsloth →

More DeepSeek models

All DeepSeek →

Other large models

All large models →