unsloth
Qwen3 VL 4B Instruct
Qwen3 VL 4B Instruct is a 4.02B 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.
Parameters4.0B
Context256K
Attentiongqa
Licenseapache-2.0
36 layers · 8 KV heads · head dim 128 · GGUF repo · config
Quant file sizes
| Quant | File size | Source |
|---|---|---|
| BF16 | 7.50 GB | Hugging Face file bytes |
| UD-Q8_K_XL | 4.71 GB | Hugging Face file bytes |
| Q8_0 | 3.99 GB | Hugging Face file bytes |
| UD-Q6_K_XL | 3.41 GB | Hugging Face file bytes |
| Q6_K | 3.08 GB | Hugging Face file bytes |
| UD-Q5_K_XL | 2.70 GB | Hugging Face file bytes |
| Q5_K_M | 2.69 GB | Hugging Face file bytes |
| Q5_K_S | 2.63 GB | Hugging Face file bytes |
| UD-Q4_K_XL | 2.37 GB | Hugging Face file bytes |
| Q4_K_M | 2.33 GB | Hugging Face file bytes |
| Q4_K_S | 2.22 GB | Hugging Face file bytes |
| IQ4_NL | 2.22 GB | Hugging Face file bytes |
| IQ4_XS | 2.11 GB | Hugging Face file bytes |
| Q4_1 | 2.42 GB | Hugging Face file bytes |
| Q4_0 | 2.21 GB | Hugging Face file bytes |
| UD-Q3_K_XL | 1.98 GB | Hugging Face file bytes |
| Q3_K_M | 1.93 GB | Hugging Face file bytes |
| Q3_K_S | 1.76 GB | Hugging Face file bytes |
| UD-IQ3_XXS | 1.56 GB | Hugging Face file bytes |
| UD-Q2_K_XL | 1.58 GB | Hugging Face file bytes |
| Q2_K | 3.11 GB | Hugging Face file bytes |
| UD-IQ2_XXS | 1.17 GB | Hugging Face file bytes |
| UD-IQ1_M | 1.06 GB | Hugging Face file bytes |
| UD-IQ1_S | 1.01 GB | Hugging Face file bytes |
On common machines
| Device | Verdict | Quant | Speed |
|---|---|---|---|
| GeForce RTX 4060 Ti 8GB | Tight fit | UD-Q8_K_XL | 25–34 |
| GeForce RTX 3060 12GB | Runs | BF16 | 21–29 |
| GeForce RTX 4060 Ti 16GB | Runs | BF16 | 17–23 |
| GeForce RTX 4090 | Runs great | BF16 | 60–82 |
| GeForce RTX 5090 | Runs great | BF16 | 106–145 |
| Apple M4 Pro | Runs great | BF16 | 16–22 |
| Apple M5 | Runs | BF16 | 9.1–12 |
| CPU only · DDR5-5600 | Runs on CPU | BF16 | 5.3–7.3 |
Smallest devices that fit it fully
- GeForce RTX 3050 6GB — Tight fit, Q8_0, 19–26 tok/s
- GeForce RTX 2060 6GB — Tight fit, Q8_0, 38–52 tok/s
- GeForce GTX 1660 Super — Tight fit, Q8_0, 38–52 tok/s
- GeForce GTX 1660 Ti — Tight fit, Q8_0, 32–44 tok/s
- Intel Arc A380 — Tight fit, Q8_0, 21–29 tok/s
- GeForce RTX 5060 Ti 8GB — Tight fit, UD-Q8_K_XL, 39–54 tok/s
Run it
Commands for the GeForce RTX 4060 Ti 16GB estimate (BF16).
llama.cpp
llama-server -hf unsloth/Qwen3-VL-4B-Instruct-GGUF:BF16 -c 8192 -ngl 36
Ollama
ollama run qwen3
LM Studio
lms get unsloth/Qwen3-VL-4B-Instruct-GGUF # Then load quant BF16