DE

DeepSeek V3 BF16

DeepSeekGGUF by unslothUpdated 2025-01-12
Download on Hugging Face
Parameters671B
Context160K
Layers61
Architecturedeepseek2
Smallest quant221.3 GB
Quick answer

DeepSeek V3 BF16 needs about 445.9 GB of memory at Q4_K_M (a 404.4 GB file) with an 8K context window, which is more than a 48 GB GPU holds. This repository has 7 quants, from 221.3 GB (Q2_K_XS) to 713.3 GB (BF16). Longer context needs more: at 32K tokens the KV cache alone adds about 163.7 GB.

DeepSeek V3 BF16 – 671B parameter Mixture‑of‑Experts language model

DeepSeek V3 BF16 is a 671B parameter Mixture‑of‑Experts model with 37B active parameters per token. It uses Multi‑head Latent Attention and DeepSeekMoE architectures. The model is released under the MIT license.

Good at

  • Large‑scale text generation
  • Efficient inference with MoE
  • Strong performance on open‑source benchmarks

Limits

  • Requires 8k context length for full capacity
  • No vision support
  • Large VRAM needed for higher quantizations

Best for

  • Chatbot back‑ends
  • Content creation pipelines
  • Research on MoE architectures

Who it’s for. Developers and researchers needing a large, open‑source language model with Mixture‑of‑Experts capabilities. The model is suitable for applications that can handle its 8k context size and require efficient inference on GPUs with sufficient VRAM.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
Q2_K_XS221.3 GB (5 parts) 262.7 GB
Too big Get file →
Q2_K_L244.2 GB (5 parts) 285.7 GB
Too big Get file →
Q3_K_M319.2 GB (7 parts) 360.6 GB
Too big Get file →
Q4_K_M404.4 GB (9 parts) 445.9 GB
Too big Get file →
Q5_K_M475.4 GB (10 parts) 516.8 GB
Too big Get file →
Q6_K550.8 GB (12 parts) 592.2 GB
Too big Get file →
BF16713.3 GB (16 parts) 754.7 GB
Too big Get file →

For 40GB VRAM use Q2_K_XS (221.25 GB file, ~262.7 GB memory at 8k). For 24GB VRAM use Q2_K_L (244.25 GB file, ~285.7 GB memory). For 16GB VRAM use Q3_K_M (319.2 GB file, ~360.6 GB memory). BF16 requires ~754.7 GB memory and 713.29 GB file.

What GPU do you need for DeepSeek V3 BF16?

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 V3 BF16 quant should you choose?

Q4_K_M is the usual starting point for DeepSeek V3 BF16: it balances file size and quality well. Here it is a 404.4 GB file.

If memory is tight, the smallest file is Q2_K_XS at 221.3 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
Parameters671B (approximate, mixture of experts)
Layers61
Trained context163,840 tokens
KV cache (fp16)4880 KB per token

Source

Quantized byunsloth
FamilyDeepSeek
Licensemit
Downloads (HF)1,527
Last updated2025-01-12

Frequently asked questions about DeepSeek V3 BF16

How much VRAM does DeepSeek V3 BF16 need?

About 445.9 GB at Q4_K_M with an 8K context window. That is the 404.4 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 V3 BF16 on an 8 GB GPU?

Not at an 8K context. The smallest file needs about 262.7 GB, so you would need a larger GPU or CPU offloading.

What is the file size of DeepSeek V3 BF16 GGUF?

It depends on the quant: Q2_K_XS is 221.3 GB, Q2_K_L is 244.2 GB, Q3_K_M is 319.2 GB, Q4_K_M is 404.4 GB, Q5_K_M is 475.4 GB, Q6_K is 550.8 GB and 1 more.

What context length does DeepSeek V3 BF16 support?

The GGUF header lists a trained context length of 163,840 tokens (160K). The KV cache grows with the context you use: about 40.9 GB at 8K and 163.7 GB at 32K.

Where can I download DeepSeek V3 BF16 GGUF?

The files are on Hugging Face in the unsloth/DeepSeek-V3-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 V3 BF16 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 V3 BF16 on 8GB VRAM?

No, the smallest quantization requires at least 40GB VRAM for offloading.

Does the model support vision tasks?

No, the model has no vision capabilities.

What license applies to this model?

The model is released under the MIT license.

Other GGUFs of DeepSeek-V3

Similar models

Browse all →

More from unsloth

All by unsloth →

More DeepSeek models

All DeepSeek →

Other large models

All large models →