Skip to content

Qwen3 VL 4B Instruct on Mac mini M4 24 GB: VRAM, speed and fit

Qwen3 VL 4B Instruct at 8-bit runs on the M4 24 GB with 72K of context at about 20 tokens a second.

Text generation · Chat · Ollama / LM Studio

Runs comfortably · on M4 24 GB · 8-bit (Q8_0) · Ollama / LM Studio · Chat

GPU memory
23.1 GB
of 24.0 GB
Context
72K
auto
Speed
20 tok/s
6.2 when full
First reply
5.8 s
2K prompt

Where the memory goes

Model
4.4 GB
Context
10.1 GB
Overhead
633 MB
OS
8.0 GB
Free
887 MB
Memory
23.1 GB of 24.0 GB
Context
72K tokens · auto
Speed
20 tok/s · 6.2 when full
First reply
5.8 s for a 2K prompt
Running cost
$0.85/mo at home · $29/mo on a cloud RTX 2000 Ada

Validation: Validated with conditions

What Qwen3 VL 4B Instruct on Mac mini M4 24 GB was checked against
CheckTargetResultStatus
Fits in GPU memory16.0 GB usable15.1 GB neededPass
Headroom5% free887 MB free (5%)Pass
Context for chat16K or more72KPass
Writing speed15 tok/s or more20 tok/sPass
First replyunder 3 s, 2K prompt5.8 sCheck

Targets: a reply you can read along with, a first reply the use case tolerates, and 5% of memory left free. Calculated from published specs, not benchmarked.

The setup

Model
Qwen3 VL 4B Instruct
Parameters
4.4B
Precision
8-bit (Q8_0)
Context
72K, auto
Chats at once
1
Use
Chat
Runs in
Ollama / LM Studio
Hardware
M4 24 GB
GPU can use
16.0 GB of 24.0 GB

Running cost

At home
$0.85/mo
39 W at the wall
Renting instead
$29/mo
RTX 2000 Ada at $0.24/h

4 h a day, power at $0.18 per kWh, cloud at RunPod on-demand prices as of Nov 22, 2026.

How to run it

llama.cpp
llama-server -hf Qwen/Qwen3-VL-4B-Instruct-GGUF
vLLM
vllm serve Qwen/Qwen3-VL-4B-Instruct

Qwen3 VL 4B Instruct on similar machines

All machines: Qwen3 VL 4B Instruct VRAM requirements. Everything that runs here: what can the Mac mini M4 24 GB run?

Other models that run on Mac mini M4 24 GB

Questions

Can Qwen3 VL 4B Instruct run on Mac mini M4 24 GB?

Qwen3 VL 4B Instruct at 8-bit runs on the M4 24 GB with 72K of context at about 20 tokens a second.

Report VR-0FSVEHK, calculated Nov 22, 2026 from published specs, not benchmarked.