DeepSeek R1 Distill Qwen 7B

• Cheapest

  NVIDIA Tesla P40

  25tok/sest.Q8_0

  $345

• Best value

  NVIDIA GeForce RTX 3080 10GB

  55tok/sest.Q8_0

  $469

• Fastest

  NVIDIA GeForce RTX 5090

  130tok/sest.Q8_0

  $4,921

• Mac

  Mac mini M4 Pro 48GB

  21tok/sest.Q8_0

  $2,149

• mlx-3bit
  • LM Studiolms get mlx-community/DeepSeek-R1-Distill-Qwen-7B-3bit
  • MLXmlx_lm.server --model mlx-community/DeepSeek-R1-Distill-Qwen-7B-3bit
• mlx-4bit
  • LM Studiolms get mlx-community/DeepSeek-R1-Distill-Qwen-7B-4bit
  • MLXmlx_lm.server --model mlx-community/DeepSeek-R1-Distill-Qwen-7B-4bit
• Q4_K_M
  • Ollamaollama run hf.co/unsloth/DeepSeek-R1-Distill-Qwen-7B-GGUF:Q4_K_M
  • LM Studiolms get unsloth/DeepSeek-R1-Distill-Qwen-7B-GGUF@q4_k_m
  • llama.cppllama-server -hf unsloth/DeepSeek-R1-Distill-Qwen-7B-GGUF:Q4_K_M
• Q5_K_M
  • Ollamaollama run hf.co/unsloth/DeepSeek-R1-Distill-Qwen-7B-GGUF:Q5_K_M
  • LM Studiolms get unsloth/DeepSeek-R1-Distill-Qwen-7B-GGUF@q5_k_m
  • llama.cppllama-server -hf unsloth/DeepSeek-R1-Distill-Qwen-7B-GGUF:Q5_K_M
• mlx-6bit
  • LM Studiolms get mlx-community/DeepSeek-R1-Distill-Qwen-7B-6bit
  • MLXmlx_lm.server --model mlx-community/DeepSeek-R1-Distill-Qwen-7B-6bit
• Q6_K
  • Ollamaollama run hf.co/unsloth/DeepSeek-R1-Distill-Qwen-7B-GGUF:Q6_K
  • LM Studiolms get unsloth/DeepSeek-R1-Distill-Qwen-7B-GGUF@q6_k
  • llama.cppllama-server -hf unsloth/DeepSeek-R1-Distill-Qwen-7B-GGUF:Q6_K
• mlx-8bit
  • LM Studiolms get mlx-community/DeepSeek-R1-Distill-Qwen-7B-8bit
  • MLXmlx_lm.server --model mlx-community/DeepSeek-R1-Distill-Qwen-7B-8bit
• Q8_0
  • Ollamaollama run hf.co/unsloth/DeepSeek-R1-Distill-Qwen-7B-GGUF:Q8_0
  • LM Studiolms get unsloth/DeepSeek-R1-Distill-Qwen-7B-GGUF@q8_0
  • llama.cppllama-server -hf unsloth/DeepSeek-R1-Distill-Qwen-7B-GGUF:Q8_0

Distilled from DeepSeek-R1 onto Qwen2.5-Math-7B (Qwen2.5 series is Apache 2.0); the card licenses the weights under MIT. Use temperature 0.5-0.7, no system prompt, and start output with "<think>\n".

How much VRAM does DeepSeek R1 Distill Qwen 7B need?

At mlx-3bit, DeepSeek R1 Distill Qwen 7B needs about 4 GB of memory at 8k context and 4 GB at 8k, counting the weights, the KV cache and runtime buffers.

What is the best GPU for running DeepSeek R1 Distill Qwen 7B locally?

The cheapest hardware in the US that runs DeepSeek R1 Distill Qwen 7B fully in memory at 8k context is the NVIDIA Tesla P40, $345 at Amazon, at about 25 tok/s (estimated). The best value is the NVIDIA GeForce RTX 3080 10GB (about 55 tok/s for $469). The fastest we price is the NVIDIA GeForce RTX 5090, at about 130 tok/s.

Can I run DeepSeek R1 Distill Qwen 7B on a 16 GB GPU?

Yes: at Q8_0 it fits fully in 16 GB at 8k context. On the NVIDIA GeForce RTX 5080 it runs at about 70 tok/s (estimated).

Can I run DeepSeek R1 Distill Qwen 7B on a 24 GB GPU?

Yes: at Q8_0 it fits fully in 24 GB at 8k context. On the NVIDIA GeForce RTX 3090 Ti it runs at about 73 tok/s (estimated).

How do I run DeepSeek R1 Distill Qwen 7B locally?

With Ollama: ollama run hf.co/unsloth/DeepSeek-R1-Distill-Qwen-7B-GGUF:Q4_K_M (the Q4_K_M build). LM Studio and llama.cpp commands are on this page.

What licence is DeepSeek R1 Distill Qwen 7B released under?

MIT License, which allows commercial use.