Can Qwen3.5 35B A3B run on Mac mini M4 64GB?

YES — Runs Great

C49Usable

▲Estimated — low-sample bucket· few comparable runs

Qwen3.5 35B A3B needs ~33.3 GB VRAM. Mac mini M4 64GB has 46.1 GB. With Q4_K_M quantization, expect ~7 tok/s.

Runtime: llama.cppCapacity: RoomyBandwidth: Very lowStack: StandardBottleneck: Memory bandwidth

Operating mode

Choose the run profile you care about

Interactive favors responsiveness, while light API and scale-out lean harder on serving readiness. The fit stays the same, but the recommendation lens changes.

Current mode

Balanced

Balanced for general local use. Keeps the ranking neutral across personal and serving workflows.

Capabilities:

Select quantization to explore

Q4_K_M (Medium quality) — 33.3 GB, 7.3 tok/s, Runs well

33.3 GB required46.1 GB available

72% VRAM used

Fit status

Runs well

Decode

7.3 tok/s

TTFT

26578 ms

Safe context

66K

Memory

33.3 GB / 46.1 GB

Memory breakdown

Weights21.3 GB

KV Cache4.1 GB

Runtime0.9 GB

Headroom6.9 GB

See how fast it feels

See how fast it feelsQwen3.5 35B A3B on Mac mini M4 64GB
1st promptCold start — includes initialization
>What is local AI inference?
Local AI inference means running an AI language model directly on your own hardware — your laptop, desktop, or server — instead of sending requests to a remote cloud API.

When you run inference locally the model weights are loaded into your GPU or unified memory. Each token you generate requires reading those weights from memory, so memory bandwidth is the main bottleneck for decode speed.

Key benefits of running locally:
 - Full privacy: your prompts never leave your machine
 - No per-token cost or rate limits
 - Works offline once the model is downloaded
 - Latency depends only on your hardware
2nd promptWarm — model ready, no init overhead
>How much VRAM do I need?
It depends on the model size and quantization level. A rough rule of thumb:

 Model size Q4 (4-bit) Q8 (8-bit) FP16
 7B params ~4.3 GB ~7.5 GB ~14 GB
 13B params ~7.9 GB ~13.9 GB ~26 GB
 70B params ~42.7 GB ~74.9 GB ~140 GB

Most people use 4-bit quantization (Q4_K_M) which gives 90-95% of full quality at a fraction of the memory. A 24 GB GPU can comfortably run most 7B-13B models.
Estimated: 7.3 tok/s decode · 26.6s TTFT (warm) · 18 tok/s prefill

What limits this setup

The model fits in shared memory, but shared-memory bandwidth is now the real limiter.

Fit does not mean dedicated-VRAM speed

Unified or shared memory can make a model technically fit, but sustained tokens per second may still trail a discrete high-bandwidth GPU with less total memory.

Shared-memory contention still exists

The OS, browser, and inference runtime all compete for the same physical memory pool, so real-world headroom is less forgiving than raw capacity suggests.

Best improvement path

Prioritize bandwidth, not only capacity

If this workload feels slow, the next useful step is often a GPU tier with materially faster memory bandwidth rather than only a small bump in capacity.

Performance by workload

Workload	Grade	Fit	Decode	TTFT	Context
Chat	C	Runs well	7.3 tok/s	14497 ms	66K
Coding	C	Runs well	7.3 tok/s	26578 ms	66K
Agentic Coding	C	Runs well	7.3 tok/s	38658 ms	66K
Reasoning	C	Runs well	7.3 tok/s	31410 ms	66K
RAG	C	Runs well	7.3 tok/s	48323 ms	66K

Quantization options

How Qwen3.5 35B A3B (35B params) fits at each quantization level on Mac mini M4 64GB (46.1 GB usable).

Quant	Bits	VRAM	Quality	Fit
Q2_K	2	13.7 GB	Low	C45
Q3_K_S	3	17.2 GB	Low	C46
NVFP4	4	19.6 GB	Medium	C47
Q4_K_M	4	21.3 GB	Medium	C47
Q5_K_M	5	25.2 GB	High	C49
Q6_K	6	28.7 GB	High	C48
Q8_0Best for your GPU	8	37.5 GB	Very High	C48
F16	16	71.8 GB	Maximum	F0

Get started

Copy-paste commands to run Qwen3.5 35B A3B on your machine.

Run

docker run --rm -it ghcr.io/ggerganov/llama.cpp:full \
 --hf-repo "lmstudio-community/Qwen3.5-35B-A3B-GGUF" \
 --hf-file "Qwen3.5-35B-A3B-GGUF-Q4_K_M.gguf" \
 -c 4096 -ngl 99

Upgrade options