The AI Bench
Alibaba (Qwen) · Apache 2.0

Qwen3-VL

Qwen’s 2025 vision models for photos, documents and screen control.

How much memory it needs

The model file, plus memory for the conversation, plus about 1 GB for the app. Longer conversations need more.

SizeVersionFileAt 8KAt 32KAt 128K
2.1BMXFP41.1 GB3.9 GB6.7 GB18.0 GB
4.4BQ4_K_M2.5 GB5.5 GB9.2 GB23.7 GB
Q8_04.3 GB7.3 GB10.9 GB25.4 GB
8.8BQ4_K_M5.0 GB8.4 GB12.0 GB26.5 GB
Q8_08.7 GB12.1 GB15.7 GB30.2 GB
31B (3B active)Q4_K_M18.6 GB21.4 GB23.9 GB33.5 GB
Q8_032.5 GB35.4 GB37.8 GB47.4 GB
33BQ4_K_M19.8 GB24.1 GB30.6 GB56.3 GB
Q8_034.8 GB39.2 GB45.6 GB71.4 GB

Which computers run it

The best size for each machine at an 8K-token conversation. Speeds are estimates from memory bandwidth.

ComputerBest versionFitSpeed
Laptop or PC, no graphics card, 32 GB8.8B Q8_0Runs great5.5–8.5tok/sReadable
RTX 4060 · 8 GB4.4B Q4_K_MRuns great45–65tok/sFast
RTX 3060 · 12 GB8.8B Q4_K_MRuns great35–55tok/sFast
RTX 5060 Ti · 16 GB8.8B Q8_0Runs well25–40tok/sSmooth
RTX 4090 · 24 GB31B Q4_K_MRuns well95–190tok/sFast
RTX 5090 · 32 GB33B Q4_K_MRuns well45–75tok/sFast
Mac with M4 · 16 GB8.8B Q4_K_MRuns great11–19tok/sSmooth
Mac with M4 Pro · 48 GB33B Q4_K_MRuns great7–12tok/sReadable
Mac with M5 Max · 128 GB33B Q8_0Runs great9.5–16tok/sReadable
Ryzen AI Max+ 395 · 128 GB33B Q4_K_MRuns great7–11tok/sReadable
DGX Spark · 128 GB33B Q4_K_MRuns great7.5–12tok/sReadable
RTX PRO 6000 · 96 GB33B Q8_0Runs great30–45tok/sSmooth
Mac Studio with M5 Ultra · 256 GB33B Q8_0Runs great15–25tok/sSmooth
Mac Studio with M3 Ultra · 512 GB33B Q8_0Runs great10–17tok/sSmooth

What it is good and bad at

Good at

  • Reading images, scanned pages and screenshots
  • Sizes from 2B to 32B

Not good at

  • Newer Qwen3.5 and Qwen3.8 also read images and are usually better

Start it

Install Ollama, then run one of these in a terminal. LM Studio users can search the same name in the app.

ollama run qwen3-vl:30b-a3b-instruct
ollama run qwen3-vl:32b-instruct