text model · DeepSeek-R1-Distill · macOS
Can I run DeepSeek-R1-0528-Qwen3-8B on Apple M4 Max (128GB)?
Yes. DeepSeek-R1-0528-Qwen3-8B runs on Apple M4 Max (128GB) at Q4_K_M (~6.2 GB of ~96 GB usable).
Runs at Q4_K_M using ~6.2 GB of ~96 GB usable. You have room for FP16 for higher quality.
That figure is at a 4k context and moves about ±15% as context length changes. Apple M4 Max (128GB) leaves ~89.8 GB of headroom, room to step up to FP16 for higher quality.
- Q4_K_M needed
- ~6.2 GB
- Usable on device
- ~96 GB
- Device memory
- 128 GB
- Best quant
- Q4_K_M
Which quant fits
Running cost · estimate
- Power draw
- ~145 W
- Electricity / 1M tokens
- ~$0.06
- Pays for itself after
- ~7,952M tok
At ~$0.15/kWh and the estimated ~93 tok/s, a million generated tokens costs about $0.06 in electricity. Versus a hosted API at ~$0.5 per million tokens, the ~$3,499 Apple M4 Max (128GB) pays for itself after roughly 7,952 million tokens, so local hardware is mostly a fixed cost, not a per-token one. TDP is the peak draw, so this is an upper bound. Assumptions.
Run it
Pick your tool. All 3 load the same Q4_K_M weights.
ollama run deepseek-r1:8b-0528-qwen3-q4_K_M llama-cli -hf unsloth/DeepSeek-R1-0528-Qwen3-8B-GGUF:Q4_K_M lms get unsloth/DeepSeek-R1-0528-Qwen3-8B-GGUF - Parameters
- 8.19B
- Q4_K_M size
- 4.68 GB
- Q8_0 size
- 8.11 GB
- Context
- 128k
- Ollama tag
- deepseek-r1:8b-0528-qwen3-q4_K_M
- Memory
- 128 GB unified
- Usable for weights
- ~96 GB
- Power draw
- ~145 W
- Best runtime
- MLX direct / Ollama (MLX backend)
You could also run
Run DeepSeek-R1-0528-Qwen3-8B on other hardware
FAQ
Can Apple M4 Max (128GB) run DeepSeek-R1-0528-Qwen3-8B?
Yes. DeepSeek-R1-0528-Qwen3-8B runs on Apple M4 Max (128GB) at Q4_K_M (~6.2 GB of ~96 GB usable).
How much memory does DeepSeek-R1-0528-Qwen3-8B need?
Apple M4 Max (128GB) has room to spare. At Q4_K_M the weights are ~4.68 GB; with KV cache and runtime overhead, budget ~6.2 GB at a 4k context.
What is the best tool to run DeepSeek-R1-0528-Qwen3-8B on macOS?
LM Studio for a simple setup; mlx-lm for the most speed. vLLM is NOT a Mac tool, it is a CUDA/Linux serving engine. Unified memory is not a fixed VRAM slice; ~70% is usable for weights.
Embed this
A live badge for your model card or README, updated as the data is. [](https://localmodel.run/can-i-run/deepseek-r1-0528-qwen3-8b/apple-m4-max-128gb) Sources
Weights are measured from GGUF files; KV cache and overhead are computed, so totals can vary ~15% with context and runtime. Any tok/s is a bandwidth estimate. See methodology.