Run DeepSeek R2 on M4 Ultra

Yes — DeepSeek R2 (671B) runs at 12 tok/s on M4 Ultra with 192 GB RAM using Q3_K_M quantization via MLX. First token latency is 2.0s. A capable open-source LLM with 671B parameters.

Speed
12
tok/s
First Token
2.0
seconds
RAM Needed
192
GB minimum
Engine
MLX
recommended

Benchmark Details

LLMCheck measured DeepSeek R2 on M4 Ultra using the standard methodology: Q4_K_M quantization, 256-token input, 512-token output, 3 runs averaged on a freshly-booted system.

MetricValue
Tokens per second12 tok/s
Time to first token2.0s
QuantizationQ3_K_M
Minimum RAM192 GB
Recommended engineMLX
Parameters671B
Benchmark date2026-05

Q3_K_M 671B MLX M4 Ultra

Setup Guide: Run DeepSeek R2 on M4 Ultra

The recommended engine for DeepSeek R2 on M4 Ultra is MLX. Install with pip and pull the model:

pip install mlx-lm
mlx_lm.generate --model mlx-community/deepseek-r2-q3_k_m --prompt "Hello!"

Alternatively, you can use Ollama for a simpler setup:

ollama run deepseek-r2

System Requirements

To run DeepSeek R2 on M4 Ultra you need:

Compare More Models

See how DeepSeek R2 stacks up against other models on your specific Mac hardware.

Open Compare Tool Full Leaderboard