unsloth
Qwen3 VL 32B Instruct
Qwen3 VL 32B Instruct is a 32.8B dense qwen3vl model from unsloth. Configured context length is 262,144 tokens. License: apache-2.0. Quant sizes are the GGUF file bytes published on Hugging Face when a file exists.
Parameters32.8B
Context256K
Attentiongqa
Licenseapache-2.0
64 layers · 8 KV heads · head dim 128 · GGUF repo · config
Quant file sizes
| Quant | File size | Source |
|---|---|---|
| BF16 | 61.0 GB | Hugging Face file bytes |
| UD-Q8_K_XL | 36.8 GB | Hugging Face file bytes |
| Q8_0 | 32.4 GB | Hugging Face file bytes |
| UD-Q6_K_XL | 27.0 GB | Hugging Face file bytes |
| Q6_K | 25.0 GB | Hugging Face file bytes |
| UD-Q5_K_XL | 21.7 GB | Hugging Face file bytes |
| Q5_K_M | 21.6 GB | Hugging Face file bytes |
| Q5_K_S | 21.1 GB | Hugging Face file bytes |
| UD-Q4_K_XL | 18.7 GB | Hugging Face file bytes |
| Q4_K_M | 18.4 GB | Hugging Face file bytes |
| Q4_K_S | 17.5 GB | Hugging Face file bytes |
| IQ4_NL | 17.4 GB | Hugging Face file bytes |
| IQ4_XS | 16.5 GB | Hugging Face file bytes |
| Q4_1 | 19.2 GB | Hugging Face file bytes |
| Q4_0 | 17.4 GB | Hugging Face file bytes |
| UD-Q3_K_XL | 15.3 GB | Hugging Face file bytes |
| Q3_K_M | 14.9 GB | Hugging Face file bytes |
| Q3_K_S | 13.4 GB | Hugging Face file bytes |
| UD-IQ3_XXS | 12.1 GB | Hugging Face file bytes |
| UD-Q2_K_XL | 11.9 GB | Hugging Face file bytes |
| Q2_K | 23.2 GB | Hugging Face file bytes |
| UD-IQ2_XXS | 8.64 GB | Hugging Face file bytes |
| UD-IQ1_M | 7.73 GB | Hugging Face file bytes |
| UD-IQ1_S | 7.21 GB | Hugging Face file bytes |
On common machines
| Device | Verdict | Quant | Speed |
|---|---|---|---|
| GeForce RTX 4060 Ti 8GB | Partial offload | Q4_K_M | 3.1–4.2 |
| GeForce RTX 3060 12GB | Tight fit | UD-IQ2_XXS | 19–26 |
| GeForce RTX 4060 Ti 16GB | Tight fit | UD-IQ3_XXS | 11–15 |
| GeForce RTX 4090 | Tight fit | UD-Q4_K_XL | 25–34 |
| GeForce RTX 5090 | Tight fit | UD-Q6_K_XL | 33–45 |
| Apple M4 Pro | Tight fit | UD-Q3_K_XL | 8.6–12 |
| Apple M5 | Tight fit | UD-IQ2_XXS | 8.1–11 |
| CPU only · DDR5-5600 | Runs on CPU | UD-Q5_K_XL | 1.9–2.6 |
Smallest devices that fit it fully
- GeForce RTX 3080 10GB — Tight fit, UD-IQ1_S, 47–65 tok/s
- Intel Arc B570 — Tight fit, UD-IQ1_S, 24–32 tok/s
- GeForce RTX 2080 Ti — Tight fit, UD-IQ1_M, 36–49 tok/s
- GeForce GTX 1080 Ti — Tight fit, UD-IQ1_M, 28–39 tok/s
- GeForce RTX 5070 — Tight fit, UD-IQ2_XXS, 36–49 tok/s
- GeForce RTX 4070 Ti — Tight fit, UD-IQ2_XXS, 27–37 tok/s
Run it
Commands for the GeForce RTX 4060 Ti 16GB estimate (UD-IQ3_XXS).
llama.cpp
llama-server -hf unsloth/Qwen3-VL-32B-Instruct-GGUF:UD-IQ3_XXS -c 4096 -ngl 64
Ollama
ollama run qwen3
LM Studio
lms get unsloth/Qwen3-VL-32B-Instruct-GGUF # Then load quant UD-IQ3_XXS