back to local model checker

Qwen3-VL 30B-A3B

MoEToolsVisionPopular

Alibaba · 31B (3B active) · Mixture of Experts

Efficient vision MoE — 3B active, strong temporal & document understanding

Apache 2.0
Downloads

0

Likes

0

Released

2025-09

Context

256K tokens

Use cases

chatvision
Total experts128
Active experts8
Active params3.0B

Your hardware

16 GB Reference Desktop

Quant
Q4_K_M
Memory
16.4 GB
Speed
~9 tok/s
Fit
Barely runs

Quantization options

What is quantization?
QuantBitsVRAMDiskQualityStatus
Q2_K210.4 GB9.5 GBlowTight fit · ~19 t/s
Q3_K_M314.4 GB13.3 GBmoderateBarely runs · ~14 t/s
Q4_K_M416.4 GB15.2 GBgoodBarely runs · ~9 t/s
Q5_K_M520.3 GB18.9 GBgoodBarely runs · ~4 t/s
Q6_K624.3 GB22.7 GBexcellentBarely runs · ~3 t/s
Q8_0832.3 GB30.3 GBexcellentBarely runs · ~2 t/s
F161664 GB60.6 GBlosslessToo heavy · ~1 t/s

About this model

Qwen3-VL 30B-A3B (Qwen3 vl 30b) is a 31B sparse mixture-of-experts (MoE) model activating 3B active per token from Alibaba in the Qwen family, designed primarily for chat, vision workflows. Supports structured tool calling; use Q4_K_M or higher when strict JSON schema adherence matters. Supports up to 256K context tokens; start at 4K–8K with Q8_0 KV cache on tight GPUs.

Hardware planning baselines start at 17.3 GB minimum memory and 28.9 GB recommended memory. At Q4_K_M, weights plus baseline runtime overhead use roughly 16.4 GB of RAM/VRAM and 15.2 GB of disk space. Because MoE routing evaluates only a subset of experts on each generated token, Qwen3-VL 30B-A3B delivers faster decoding throughput per gigabyte of memory bandwidth than a dense checkpoint of identical total size—provided the full 16.4 GB checkpoint stays resident without SSD paging.

Deploy Qwen3-VL 30B-A3B locally via Ollama (`qwen3-vl:30b-a3b`), LM Studio, or Ollama with Flash Attention enabled. Under Apache License 2.0, review official weight cards before commercial redistribution.

Highlights

  • Local Fit Baseline17.3 GB minimum and 28.9 GB recommended RAM/VRAM.
  • Primary QuantQ4_K_M targets 16.4 GB memory and 15.2 GB disk.
  • Context & Capabilities256K max context · Sparse MoE · Tool calling.

System requirements by device

Memory planning targets 17.3 GB minimum RAM/VRAM and 28.9 GB recommended memory. The baseline quant is Q4_K_M, estimated at 16.4 GB memory and 15.2 GB disk. Actual runtime usage shifts with context length, KV cache precision, and OS background overhead.

DeviceNotes
Mac (Apple silicon)Use unified memory headroom (~75% physical RAM max) as the primary fit signal. Start with Q4_K_M and close heavy browser tabs first.
Windows / Linux GPUKeep weights plus KV cache inside dedicated VRAM so Windows WDDM does not spill layers over PCIe into Shared GPU Memory.
Phones & Compact Edge BoardsNot practical on phones (17.3 GB min RAM exceeds Android LMK and iOS Jetsam limits). Pick a 2B–4B model on mobile.

Start with a short session and leave memory for the operating system. A file fitting on disk does not guarantee that it will load in memory.

Estimated memory by quantization

These are rough estimates, not tested minimum requirements. Check the actual file and runtime documentation; context, offloading and operating-system overhead can change the memory needed.

Quant TierEstimated memoryDisk SizeQuality
Q2_K10.4 GB9.5 GBlow
Q3_K_M14.4 GB13.3 GBmoderate
Q4_K_M16.4 GB15.2 GBgood
Q5_K_M20.3 GB18.9 GBgood
Q6_K24.3 GB22.7 GBexcellent
Q8_032.3 GB30.3 GBexcellent
F1664.0 GB60.6 GBlossless

How to run Qwen3-VL 30B-A3B on your device

Start with a short prompt and the smallest quant that still gives the quality you need. A clean first run is more useful than loading a larger file with no memory headroom.

Mac

Run Qwen3-VL 30B-A3B on Mac

  1. Install Ollama or LM Studio on macOS Apple Silicon.
  2. Pull Q4_K_M (16.4 GB) first and confirm it stays below ~75% of your Mac's unified memory.
  3. Enable Flash Attention and cap initial context at 8K until memory pressure stays below 28.9 GB.
  4. Consult the Mac guide for Activity Monitor tuning and Metal offload tips.
Device how-to guide
Windows

Run Qwen3-VL 30B-A3B on Windows

  1. Install LM Studio for Windows for visual VRAM inspection, or Ollama when you want a localhost API endpoint.
  2. Load Q4_K_M (~16.4 GB) and verify Dedicated GPU Memory stays within VRAM without spilling into Shared GPU Memory.
  3. Reserve at least 17.3 GB memory minimum and target 28.9 GB for comfortable context growth.
  4. Enable Q8_0 KV cache or drop one quant tier if generation slows down over PCIe.
Device how-to guide

Qwen3-VL 30B-A3B vs alternatives

ModelParamsContextMin RAMLicense
Qwen3-VL 30B-A3B31B (3B active)256K17.3 GBApache 2.0
Qwen 3 30B-A3B30B (3.3B active)128K16.8 GBApache 2.0
Qwen 2.5 32B32B128K17.9 GBApache 2.0
Qwen 2.5 Coder 32B32B128K17.9 GBApache 2.0
Qwen 3 32B32B128K17.9 GBApache 2.0
Qwen 3 Coder 30B-A3B30B (3B active)256K16.8 GBApache 2.0

Use this table to choose the smallest nearby model that still matches the job. A related model with lower RAM can be the better local choice when the current page fits only at the edge of available memory.

Frequently asked questions

What is Qwen3-VL 30B-A3B?
Qwen3-VL 30B-A3B is a 31B sparse mixture-of-experts model (3B active active per token) from Alibaba, supporting up to 256K tokens with 17.3 GB minimum RAM and 28.9 GB recommended RAM.
How do I download and install Qwen3 vl 30b?
Download a GGUF build for Qwen3-VL 30B-A3B, start with Q4_K_M, and confirm the model fits locally before increasing context or quant quality.
How much VRAM or RAM does Qwen3-VL 30B-A3B require?
At Q4_K_M, Qwen3-VL 30B-A3B occupies roughly 16.4 GB of active memory and 15.2 GB on disk. Add 2–6 GB of spare headroom for KV cache buffers, runtime scratch space, and OS apps (28.9 GB recommended).
What is the best way to run Qwen3-VL 30B-A3B locally?
Use `ollama run qwen3-vl:30b-a3b` for a quick CLI/API setup, or load Q4_K_M in LM Studio or llama.cpp with Flash Attention enabled when you want explicit KV cache and layer control.
Can Qwen3-VL 30B-A3B run on Android or iPhone?
Qwen3-VL 30B-A3B requires 17.3 GB minimum memory, placing it outside practical 8 GB smartphone limits. Run it on a 16 GB+ Mac or desktop GPU instead, and pick a 2B–4B checkpoint for phones.
How does Qwen3-VL 30B-A3B compare to Qwen 3 30B-A3B?
Qwen 3 30B-A3B (30B, 16.8 GB min RAM) sits closest in size. Compare active parameters, context ceiling (256K vs 128K), and license terms (Apache 2.0) before switching.