[whatmodel.sh](../../README.md) / Qwen3-VL 4B Instruct
# Qwen3-VL 4B Instruct
Alibaba · released 2025-10 · 4.44B params · apache-2.0
**score 52.8** ±0.7 · #322 of 732 · from 27 benchmark results across 3 sources
Score within the [LMArena (text)](https://whatmodel.sh/?set=lmarena-text&score=set) comparison set only.
## What it's good at
● marks Qwen3-VL 4B Instruct among every model on that task or test, worst to best.
task vs the others rank counts tests
agent tasks ▁▁▃▆●▄▅▅▄▄▂▁ #211 of 313* 2.1% 1
coding ▁▁▁▃▅●█▇▅▄▃▁ #344 of 469* 3.9% 1
chat & writing ▁▁▁▃▆▄▆●▆▇▇▃ #211 of 402* 3.1% 1
math ▁▁▃▅█●▄█▄▂▁▁ #207 of 431* 10% 2
vision & documents ▁▁▂▂▁●▄▂▆▇█▄ #119 of 153 51% 15
agent tasks · #211 of 313 · 2.1% of its score
Beats [o1 (2024-12-17)](o1-2024-12-17.md), [o3-mini](o3-mini.md), [DeepSeek R1 (Jan '25)](deepseek-r1.md) here, though they rank higher overall. Trails [SmolLM3 3B](smollm3-3b.md), [Jamba 1.5 Mini](jamba-1-5-mini.md), [Mellum2 12B-A2.5B Instruct](mellum2-12b-a2-5b-instruct.md) here, though they rank lower overall.
test result vs the others in its run difficulty counts why
BFCL v3 (maker) 63.3% █▄··▄●·█▄█▄█ beats 5/13 medium · avg 59% 2.1% blunt · at its level
coding · #344 of 469 · 3.9% of its score
Trails [Phi-3-mini-128k-instruct (Apr 2024)](phi-3-mini-128k-instruct.md), [DeepSeek LLM 67B Chat](deepseek-llm-67b-chat.md), [WizardLM 70B V1.0](wizardlm-70b-v1-0.md) here, though they rank lower overall.
test result vs the others in its run difficulty counts why
LiveCodeBench v6 (window unspecif… 37.9% ▃▅●▅▅▃▅···▃█ beats 4/16 hard · avg 35% 3.9% at its level
chat & writing · #211 of 402 · 3.1% of its score
Beats [Gemma 2 9B IT SimPO](gemma-2-9b-simpo.md), [Granite 4.2 8B](granite-4-2-8b.md), [Mistral Small 4 (119B-A6.5B)](mistral-small-4.md) here, though they rank higher overall. Trails [LFM2.5 1.2B Instruct](lfm2-5-1-2b-instruct.md), [Gemma 3n E4B](gemma-3n-e4b.md), [LFM2.5 1.2B Thinking](lfm2-5-1-2b-thinking.md) here, though they rank lower overall.
test result vs the others in its run difficulty counts why
IFEval (aggregation unspecified) … 82.3% ▃··▃···●█▅▃▃ beats 3/11 easy · avg 77% 3.1% typical
math · #207 of 431 · 10% of its score
Beats [MiniMax-M2.1](minimax-m2-1.md), [DeepSeek V3 (Dec '24)](deepseek-v3.md), [Qwen3 235B-A22B (Apr 2025, hybrid)](qwen3-235b-a22b.md) here, though they rank higher overall. Trails [Granite 8B Code Base (4K)](granite-8b-code-base-4k.md), [Mixtral 8x7B v0.1 (base)](mixtral-8x7b-v0-1.md), [Olmo 3 7B Instruct](olmo-3-7b-instruct.md) here, though they rank lower overall.
test result vs the others in its run difficulty counts why
MathVista (testmini) (maker) 73.7% ▃··▃▅·▃●▅▅▅█ beats 7/16 medium · avg 69% 5.5% at its level
AIME 2025 (maker) 46.6% ▃·▃▃●··▃▅▃·█ beats 4/12 medium · avg 44% 4.8% sharp · at its level
vision & documents · #119 of 153 · 51% of its score
Beats [DeepSeek V4 Flash Vision (Exp)](deepseek-v4-flash-vision-exp.md), [GPT-4 Turbo (2024-04-09)](gpt-4-turbo-2024-04-09.md), [GPT-4o mini (2024-07-18)](gpt-4o-mini-2024-07-18.md) here, though they rank higher overall.
test result vs the others in its run difficulty counts why
CharXiv (reasoning questions) (ma… 39.7% ▄▂●▆▄▄▆··▄█▂ beats 3/21 medium · avg 38% 9.2% at its level
MMMU (val) (maker) 67.4% ▂···▆▂▂▂●▄▄█ beats 7/16 medium · avg 61% 7.6% at its level
CharXiv descriptive questions (ma… 76.2% ▄··▂▂●·▂▄▄▂█ beats 4/14 easy · avg 74% 7.2% typical
MathVista (testmini) (maker) 73.7% ▃··▃▅·▃●▅▅▅█ beats 7/16 medium · avg 69% 5.5% at its level
MMStar (maker) 69.8% ▂··▄▂▂·▄●▂▄▆ beats 7/16 medium · avg 65% 5.1% at its level
MMMU Pro (maker) 53.2% ▃·█·▅▃▃●▅▅▃█ beats 8/16 medium · avg 46% 4.2% at its level
ChartArena chart parsing (English… 33.2% ▁·▁·▃▃▁●▃█·▄ beats 7/22 hard · avg 27% 2.4% blunt · at its level
HallusionBench (maker) 57.6% ▃·▃·▅▃▅●▃▅█▅ beats 7/16 medium · avg 54% 2.0% blunt · at its level
ChartQA (maker) 84.6% ▂··▂·▂●·▅▂▆█ beats 4/17 easy · avg 85% 1.6% blunt
RealWorldQA (maker) 70.9% ▅▃·▃▅●·█·▃▅█ beats 6/16 easy · avg 70% 1.5% blunt · at its level
DocVQA (maker) 95.3% ▃··▅▃▅▃●▃█▅▅ beats 8/17 very easy · avg 94% 1.5% blunt · maxed out
ScreenSpot-Pro (maker) 59.5% ▂····▂▄··▂●▆ beats 8/11 medium · avg 46% 1.2% blunt · at its level
CC-OCR (maker) 76.2% ▃··▃▅·●▅·█▅▃ beats 4/13 easy · avg 74% 1.0% blunt
OCRBench (maker) 88.1% ▅█·▃▃█▃●▅▃·▃ beats 11/16 easy · avg 84% 0.5% blunt
PureDocBench document parsing (Av… 67.5% ▂·▁·▂▂●█·▁▂▂ beats 9/22 medium · avg 62% 0.1% blunt · at its level
other tests (8)
test result vs the others in its run difficulty counts why
GPQA (subset unspecified) (maker) 55.9% ▂····▂·▂●█▆█ beats 3/14 hard · avg 49% 8.1% sharp · at its level
MMLU-Pro (maker) 67.1% ▄·▄·▄●▄▄█▄██ beats 3/12 medium · avg 63% 7.5% at its level
MMLU-Redux (maker) 81.5% ▃··▃·●·█▃▅▃▅ beats 2/11 easy · avg 79% 6.0% typical
SuperGPQA (maker) 40.3% ▄··▄·●▄▄█▄██ beats 2/11 medium · avg 37% 3.9% at its level
AndroidWorld (maker) 45.3% ▃··●▅▃▃▃▃▃·█ beats 1/11 medium · avg 43% 3.0% at its level
Video-MME (without subtitles) (ma… 69.3% ▃··█··●█▃▃▅▃ beats 5/13 medium · avg 66% 2.8% blunt · at its level
LVBench (maker) 56.2% ▅··▃·●▃▃·█·▃ beats 4/10 hard · avg 51% 2.7% at its level
MVBench (maker) 68.9% ▂▂▂·▂●··▆▄·▂ beats 5/13 medium · avg 67% 1.1% blunt · at its level
source results share of the score
Qwen (submitted / maker-reported) 25 97% ▇▇▇▇▇▇▇▇▇▇▇▇
ChartArena: Benchmarking Chart Parsing … 1 2% ▇
How Far Is Document Parsing from Solved… 1 0% ▇
prior: pull toward the middle 0% ▇
weighted average = 52.8
Licences: ChartArena: Benchmarking Chart Parsing across Languages ([CC BY 4.0](https://creativecommons.org/licenses/by/4.0/)) · How Far Is Document Parsing from Solved? PureDocBench: A Source-Traceable Benchmark across Clean ([CC BY 4.0](https://creativecommons.org/licenses/by/4.0/))
## How to run it
### llama.cpp (recommended)
1. Install llama.cpp. Homebrew and winget installs name the commands llama-cli and llama-server instead. [Other ways to install](https://github.com/ggml-org/llama.cpp/blob/master/docs/install.md)
```sh
# macOS / Linux
curl -LsSf https://llama.app/install.sh | sh
# Windows
irm https://llama.app/install.ps1 | iex
```
2. Run Qwen3-VL 4B Instruct (Q4_K_M, needs ~2.5 GB, from [huggingface.co/Qwen](https://huggingface.co/Qwen/Qwen3-VL-4B-Instruct-GGUF/tree/1cd86afb9a95c410a6038ab3b40d8b578c892266))
```sh
llama cli -hf Qwen/Qwen3-VL-4B-Instruct-GGUF:Q4_K_M
```
3. Or start a local server and open the address it prints (http://localhost:9931)
```sh
llama serve -hf Qwen/Qwen3-VL-4B-Instruct-GGUF:Q4_K_M
```
Other versions: Q8_0 (~4.3 GB) `llama cli -hf Qwen/Qwen3-VL-4B-Instruct-GGUF:Q8_0`
### LM Studio
1. Install LM Studio. Open the app once before using the lms command. [Download it (macOS / Windows / Linux)](https://lmstudio.ai/download)
2. Run Qwen3-VL 4B Instruct (Q4_K_M, needs ~2.5 GB, from [huggingface.co/Qwen](https://huggingface.co/Qwen/Qwen3-VL-4B-Instruct-GGUF/tree/1cd86afb9a95c410a6038ab3b40d8b578c892266))
```sh
lms get https://huggingface.co/Qwen/Qwen3-VL-4B-Instruct-GGUF@Q4_K_M -y
```
3. Then run lms ls to see its name, and lms chat followed by that name
Other versions: Q8_0 (~4.3 GB) `lms get https://huggingface.co/Qwen/Qwen3-VL-4B-Instruct-GGUF@Q8_0 -y` · 4bit (~3.1 GB, Mac only) `lms get https://huggingface.co/lmstudio-community/Qwen3-VL-4B-Instruct-MLX-4bit -y` · 8bit (~5.1 GB, Mac only) `lms get https://huggingface.co/lmstudio-community/Qwen3-VL-4B-Instruct-MLX-8bit -y`
### Ollama
1. Install Ollama. [Download it (macOS / Windows)](https://ollama.com/download)
```sh
# Linux
curl -fsSL https://ollama.com/install.sh | sh
```
2. Run Qwen3-VL 4B Instruct (Q4_K_M, needs ~3.3 GB, from [ollama.com](https://ollama.com/library/qwen3-vl/tags))
```sh
ollama run qwen3-vl:4b-instruct
```
Other versions: Q8_0 (~5.1 GB) `ollama run qwen3-vl:4b-instruct-q8_0`
### MLX (Mac)
1. Install MLX (Mac). Apple silicon Macs only, text only.
```sh
# with uv
uv tool install mlx-lm
# or
pip install mlx-lm
```
2. Run Qwen3-VL 4B Instruct (4bit, needs ~3.1 GB, Mac only, from [huggingface.co/lmstudio-community](https://huggingface.co/lmstudio-community/Qwen3-VL-4B-Instruct-MLX-4bit/tree/552af30c9952c44f1e1a27c7c5810ded58e892bc))
```sh
mlx_lm.chat --model lmstudio-community/Qwen3-VL-4B-Instruct-MLX-4bit
```
3. Or start an OpenAI-compatible server at http://localhost:8080/v1
```sh
mlx_lm.server --model lmstudio-community/Qwen3-VL-4B-Instruct-MLX-4bit
```
Other versions: 8bit (~5.1 GB, Mac only) `mlx_lm.chat --model lmstudio-community/Qwen3-VL-4B-Instruct-MLX-8bit`
Model repository and documentation: [huggingface.co/Qwen](https://huggingface.co/Qwen/Qwen3-VL-4B-Instruct)
[interactive version](https://whatmodel.sh/model/qwen3-vl-4b-instruct) · [JSON](https://whatmodel.sh/api?format=json&all=1)