text model · Qwen3-Coder · macOS
Can I run Qwen3-Coder 30B-A3B on Apple M3 Ultra (256GB)?
Yes. Qwen3-Coder 30B-A3B runs on Apple M3 Ultra (256GB) at Q4_K_M (~19.4 GB of ~192 GB usable).
Runs at Q4_K_M using ~19.4 GB of ~192 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 M3 Ultra (256GB) leaves ~172.6 GB of headroom, room to step up to FP16 for higher quality.
- Q4_K_M needed
- ~19.4 GB
- Usable on device
- ~192 GB
- Device memory
- 256 GB
- Best quant
- Q4_K_M
Which quant fits
Run it
Pick your tool. All 3 load the same Q4_K_M weights.
ollama run qwen3-coder:30b llama-cli -hf unsloth/Qwen3-Coder-30B-A3B-Instruct-GGUF:Q4_K_M lms get unsloth/Qwen3-Coder-30B-A3B-Instruct-GGUF - Parameters
- 30.5B (MoE, 3.3B active)
- Q4_K_M size
- 17.28 GB
- Q8_0 size
- 30.25 GB
- Context
- 256k
- Ollama tag
- qwen3-coder:30b
- Memory
- 256 GB unified
- Usable for weights
- ~192 GB
- Power draw
- ~270 W
- Best runtime
- MLX direct / Ollama (MLX backend)
You could also run
Run Qwen3-Coder 30B-A3B on other hardware
FAQ
Can Apple M3 Ultra (256GB) run Qwen3-Coder 30B-A3B?
Yes. Qwen3-Coder 30B-A3B runs on Apple M3 Ultra (256GB) at Q4_K_M (~19.4 GB of ~192 GB usable).
How much memory does Qwen3-Coder 30B-A3B need?
Apple M3 Ultra (256GB) has room to spare. At Q4_K_M the weights are ~17.28 GB; with KV cache and runtime overhead, budget ~19.4 GB at a 4k context. It is a Mixture-of-Experts model (30.5B total / 3.3B active), so all experts must stay in memory; memory tracks total params, not active params.
What is the best tool to run Qwen3-Coder 30B-A3B 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/qwen3-coder-30b-a3b/apple-m3-ultra-256gb) 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.