Yes — DeepSeek R1 32B (32B) runs at 14 tok/s on M3 Max with 36 GB RAM using Q4_K_M quantization via Ollama. First token latency is 2.0s. DeepSeek's 32B reasoning model delivering frontier-grade results locally.
LLMCheck measured DeepSeek R1 32B on M3 Max using the standard methodology: Q4_K_M quantization, 256-token input, 512-token output, 3 runs averaged on a freshly-booted system.
| Metric | Value |
|---|---|
| Tokens per second | 14 tok/s |
| Time to first token | 2.0s |
| Quantization | Q4_K_M |
| Minimum RAM | 36 GB |
| Recommended engine | Ollama |
| Parameters | 32B |
| Benchmark date | 2025-12 |
Q4_K_M 32B Ollama M3 Max
The recommended engine for DeepSeek R1 32B on M3 Max is Ollama. Install Ollama, then pull the model:
Ollama handles quantization automatically — it will download the Q4_K_M variant (~36 GB) and start an interactive chat session.
| Chip | Speed | First Token | Min RAM | Engine |
|---|---|---|---|---|
| M5 Max | 27 tok/s | 1.2s | 64 GB | Ollama |
| M4 Max | 18 tok/s | 1.8s | 48 GB | LM Studio |
To run DeepSeek R1 32B on M3 Max you need:
See how DeepSeek R1 32B stacks up against other models on your specific Mac hardware.
Open Compare Tool Full Leaderboard