DE

deepseek llm 67b base

DeepSeekGGUF by TheBlokeUpdated 2023-11-29
Download on Hugging Face
Parameters67.4B
Context4K
Layers95
Architecturellama
Smallest quant28.5 GB
Quick answer

deepseek llm 67b base needs about 44.1 GB of memory at Q4_K_M (a 40.4 GB file) with an 8K context window, so it fits on a 48 GB GPU. This repository has 10 quants, from 28.5 GB (Q2_K) to 47.7 GB (Q5_K_M).

DeepSeek LLM 67B Base – a large language model in GGUF format

DeepSeek LLM 67B Base is a 67.4 billion‑parameter model trained from scratch on 2 trillion tokens. It uses Grouped‑Query Attention and supports a 4096‑token context window. The model is available in multiple GGUF quantisations for local deployment.

Good at

  • Large‑scale text generation
  • Multilingual support (English & Chinese)
  • Research‑grade model for fine‑tuning

Limits

  • No vision support
  • Requires significant VRAM for higher quants
  • License is MIT for code, model license applies

Best for

  • Academic research on large‑language models
  • Fine‑tuning for domain‑specific applications
  • Benchmarking new inference engines

Who it’s for. Researchers and developers who need a 67 billion‑parameter model for experimentation or fine‑tuning. The model is open source and supports commercial use under the model license.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
Q2_K28.5 GB 30.6 GB
Too big Get file →
Q3_K_S29.3 GB 31.4 GB
Too big Get file →
Q3_K_M32.6 GB 34.7 GB
Too big Get file →
Q3_K_L35.6 GB 37.7 GB
Too big Get file →
Q4_038.2 GB 40.2 GB
Too big Get file →
Q4_K_S38.2 GB 40.3 GB
Too big Get file →
Q4_K_M40.4 GB 42.5 GB
Too big Get file →
Q5_046.5 GB 48.6 GB
Too big Get file →
Q5_K_S46.5 GB 48.6 GB
Too big Get file →
Q5_K_M47.7 GB 49.7 GB
Too big Get file →

For 8 GB VRAM, use Q2_K (28.5 GB file, ~32.2 GB memory at 8k context). For 12 GB VRAM, Q3_K_S (29.31 GB file, ~33 GB memory) is suitable. If you have 16 GB or more, consider Q4_0 (38.15 GB file, ~41.8 GB memory) for better quality.

What GPU do you need for deepseek llm 67b base?

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: the best quant that fits is Q4_K_M (40.4 GB file, about 44.1 GB in use).

Which deepseek llm 67b base quant should you choose?

Q4_K_M is the usual starting point for deepseek llm 67b base: it balances file size and quality well. Here it is a 40.4 GB file.

If memory is tight, the smallest file is Q2_K at 28.5 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

Architecturellama
Parameters67.4B
Layers95
Trained context4,096 tokens
KV cache (fp16)380 KB per token

Source

Quantized byTheBloke
FamilyDeepSeek
Licenseother
Downloads (HF)1,765
Last updated2023-11-29

Frequently asked questions about deepseek llm 67b base

How much VRAM does deepseek llm 67b base need?

About 44.1 GB at Q4_K_M with an 8K context window. That is the 40.4 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 llm 67b base on an 8 GB GPU?

Not at an 8K context. The smallest file needs about 32.2 GB, so you would need a larger GPU or CPU offloading.

What is the file size of deepseek llm 67b base GGUF?

It depends on the quant: Q2_K is 28.5 GB, Q3_K_S is 29.3 GB, Q3_K_M is 32.6 GB, Q3_K_L is 35.6 GB, Q4_0 is 38.2 GB, Q4_K_S is 38.2 GB and 4 more.

What context length does deepseek llm 67b base support?

The GGUF header lists a trained context length of 4,096 tokens (4K). The KV cache grows with the context you use: about 3.2 GB at 8K and 1.6 GB at 4K.

Where can I download deepseek llm 67b base GGUF?

The files are on Hugging Face in the TheBloke/deepseek-llm-67b-base-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 llm 67b base 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 LLM 67B Base on 8 GB VRAM?

Only the Q2_K quant can fit, using about 32.2 GB of memory at 8k context, which exceeds 8 GB VRAM. You would need a higher‑capacity GPU.

Does the model support Chinese text?

Yes, the training data includes Chinese, so the model can generate and understand Chinese text.

Is the model licensed for commercial use?

The model code is MIT licensed, and the model itself supports commercial use under its model license.

Other GGUFs of deepseek-llm-67b-base

Similar models

Browse all →

More from TheBloke

All by TheBloke →

More DeepSeek models

All DeepSeek →

Other large models

All large models →