Qwen3.5-35B-A3B (MoE) on Apple Mac mini M5 Pro (64GB Unified): VRAM Requirement & Speed Estimate
Yes. At Q4_K_M 4-bit quantization with a 16K context window, Qwen3.5-35B-A3B (MoE) requires 23.34 GB of VRAM (21.22 GB weights + 0.4 GB KV cache + 1.72 GB overhead), fitting within Mac mini M5 Pro 64G's 52.5 GB usable memory at ~100.3 tokens/s.
Physical 64 GB · Load 44.5%
MoE: reads only 1.82 GB active weights/step
Enable Q8/Q4 KV quantization to scale batch concurrency
OLLAMA_CONTEXT_LENGTH=16384 ollama run qwen3.5:35b-a3bMac mini M5 Pro 64G Fits Qwen3.5-35B-A3B (MoE) · Cost & TCO
Approx. $1,999 (~$31/GB VRAM) · 65W TDP
📊 All Quantization Levels for Qwen3.5-35B-A3B (MoE) on Mac mini M5 Pro 64G
| Quantization | Total VRAM (16K) | Verdict | Est. Speed | Max Context |
|---|---|---|---|---|
| FP16 / BF16 | 73.1 GB | Out of VRAM (OOM) | — | — |
| FP8 / Q8_0 | 39.63 GB | Smooth Fit | ~59.8 tok/s | 256K |
| GGUF Q6_K | 30.98 GB | Smooth Fit | ~76.1 tok/s | 256K |
| GGUF Q5_K_M | 27.09 GB | Smooth Fit | ~86.8 tok/s | 256K |
| Q4_K_M / AWQ | 23.34 GB | Smooth Fit | ~100.3 tok/s | 256K |
| GGUF IQ4_XS | 20.67 GB | Smooth Fit | ~112.9 tok/s | 256K |
| GGUF Q3_K_M | 19.15 GB | Smooth Fit | ~121.5 tok/s | 256K |
| GGUF Q2_K | 15.76 GB | Smooth Fit | ~146.6 tok/s | 256K |
💡 Recommended Alternative GPUs & Macs for Qwen3.5-35B-A3B (MoE)
Frequently Asked Questions
Can Apple Mac mini M5 Pro (64GB Unified) run Qwen3.5-35B-A3B (MoE) locally?
Yes. At Q4_K_M 4-bit quantization with a 16K context window, Qwen3.5-35B-A3B (MoE) requires 23.34 GB of VRAM (21.22 GB weights + 0.4 GB KV cache + 1.72 GB overhead), fitting within Mac mini M5 Pro 64G's 52.5 GB usable memory at ~100.3 tokens/s.
What is the best quantization for Qwen3.5-35B-A3B (MoE) on Mac mini M5 Pro 64G?
The highest-quality quantization that fits cleanly on Mac mini M5 Pro 64G is FP8 / Q8_0 (99.9% Near-Lossless), using 39.63 GB VRAM at ~59.8 tokens/s with up to 256K context.
How much VRAM does Qwen3.5-35B-A3B (MoE) need in full FP16 precision?
In unquantized FP16/BF16 precision (16K context), Qwen3.5-35B-A3B (MoE) requires 73.1 GB VRAM (70 GB for weights alone). Because it uses a Mixture-of-Experts (MoE) architecture, all 35B parameters must sit in memory, while only 3B active parameters are read per token.