[whatmodel.sh](../../README.md) / Qwen2.5-VL 3B Instruct
# Qwen2.5-VL 3B Instruct
Alibaba · released 2025-01 · 3.75B params · qwen-research
**score 44.6** ±1.0 · #485 of 732 · from 14 benchmark results across 2 sources
Score within the [LMArena (text)](https://whatmodel.sh/?set=lmarena-text&score=set) comparison set only.
## What it's good at
● marks Qwen2.5-VL 3B Instruct among every model on that task or test, worst to best.
task vs the others rank counts tests
math ▁▁▃▅●▆▄█▄▂▁▁ #299 of 431* 12% 1
vision & documents ▁▁●▂▁▃▄▂▆▇█▄ #139 of 153 100% 14
math · #299 of 431 · 12% of its score
Beats [Molmo 2 8B](molmo-2-8b.md), [DeepSeek-Coder-V2 Instruct (0724)](deepseek-coder-v2-0724.md), [Llama 3.1 Nemotron 70B Instruct](llama-3-1-nemotron-70b.md) here, though they rank higher overall. Trails [DeepSeek R1 Distill Qwen 1.5B](deepseek-r1-distill-qwen-1-5b.md), [Gemma 7B (base)](gemma-7b-base.md), [Mistral 7B v0.1 (base)](mistral-7b-v0-1.md) here, though they rank lower overall.
test result vs the others in its run difficulty counts why
MathVista (testmini) (maker) 62.3% ▃··▃●·▃█▅▅▅█ beats 3/16 medium · avg 69% 12% at its level
vision & documents · #139 of 153 · 100% of its score
Beats [Gemma 3 4B](gemma-3-4b.md), [LLaVA-OneVision Qwen2 72B (ov-sft)](llava-onevision-qwen2-72b-ov.md) here, though they rank higher overall. Trails [Qwen3-VL 2B Instruct](qwen3-vl-2b-instruct.md) here, though they rank lower overall.
test result vs the others in its run difficulty counts why
CharXiv descriptive questions (ma… 58.6% ●··▂▂▂·▂▄▄▂█ beats 0/14 easy · avg 74% 19% at its level
CharXiv (reasoning questions) (ma… 31.3% ●▂▄▆▄▄▆··▄█▂ beats 1/21 medium · avg 38% 15% at its level
MMMU (val) (maker) 53.1% ▂···●▂▂▂▄▄▄█ beats 1/16 medium · avg 61% 14% at its level
MathVista (testmini) (maker) 62.3% ▃··▃●·▃█▅▅▅█ beats 3/16 medium · avg 69% 12% at its level
MMStar (maker) 55.9% ▂··●▂▂·▄█▂▄▆ beats 1/16 medium · avg 65% 10% at its level
MMMU Pro (maker) 31.6% ●·█·▅▃▃▃▅▅▃█ beats 0/16 medium · avg 46% 6.8% at its level
HallusionBench (maker) 46.3% ▃·●·▅▃▅▅▃▅█▅ beats 1/16 medium · avg 54% 4.1% blunt · at its level
ChartQA (maker) 84.0% ▂··▂·▂●·▅▂▆█ beats 3/17 easy · avg 85% 4.0% blunt
DocVQA (maker) 93.9% ▃··▅●▅▃█▃█▅▅ beats 3/17 very easy · avg 94% 3.9% blunt · maxed out
ChartArena chart parsing (English… 21.4% ▁·▁·●▃▁▇▃█·▄ beats 2/22 hard · avg 27% 3.7% blunt
RealWorldQA (maker) 65.4% ▅●·▃▅▅·█·▃▅█ beats 2/16 easy · avg 70% 3.2% blunt · at its level
CC-OCR (maker) 74.5% ▃··▃●·▅▅·█▅▃ beats 3/13 easy · avg 74% 2.2% blunt · at its level
OCRBench (maker) 79.7% ●█·▃▃█▃▅▅▃·▃ beats 1/16 easy · avg 84% 1.2% blunt
TextVQA (maker) 79.3% ●······▄·▄▄▄ beats 0/5 easy · avg 80% 0.7% blunt
source results share of the score
Qwen (submitted / maker-reported) 13 96% ▇▇▇▇▇▇▇▇▇▇▇▇
ChartArena: Benchmarking Chart Parsing … 1 4% ▇
prior: pull toward the middle 0% ▇
weighted average = 44.6
Licences: ChartArena: Benchmarking Chart Parsing across Languages ([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 Qwen2.5-VL 3B Instruct (Q4_K_M, needs ~1.9 GB, from [huggingface.co/ggml-org](https://huggingface.co/ggml-org/Qwen2.5-VL-3B-Instruct-GGUF/tree/5037fcf163dd95d1e41d1974465f0898ed108ca2))
```sh
llama cli -hf ggml-org/Qwen2.5-VL-3B-Instruct-GGUF:Q4_K_M
```
3. Or start a local server and open the address it prints (http://localhost:9931)
```sh
llama serve -hf ggml-org/Qwen2.5-VL-3B-Instruct-GGUF:Q4_K_M
```
Other versions: Q8_0 (~3.3 GB) `llama cli -hf ggml-org/Qwen2.5-VL-3B-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 Qwen2.5-VL 3B Instruct (Q4_K_M, needs ~1.9 GB, from [huggingface.co/ggml-org](https://huggingface.co/ggml-org/Qwen2.5-VL-3B-Instruct-GGUF/tree/5037fcf163dd95d1e41d1974465f0898ed108ca2))
```sh
lms get https://huggingface.co/ggml-org/Qwen2.5-VL-3B-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 (~3.3 GB) `lms get https://huggingface.co/ggml-org/Qwen2.5-VL-3B-Instruct-GGUF@Q8_0 -y` · 4bit (~3.1 GB, Mac only) `lms get https://huggingface.co/mlx-community/Qwen2.5-VL-3B-Instruct-4bit -y` · 8bit (~4.1 GB, Mac only) `lms get https://huggingface.co/mlx-community/Qwen2.5-VL-3B-Instruct-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 Qwen2.5-VL 3B Instruct (Q4_K_M, needs ~3.2 GB, from [ollama.com](https://ollama.com/library/qwen2.5vl/tags))
```sh
ollama run qwen2.5vl:3b-q4_K_M
```
Other versions: Q8_0 (~4.6 GB) `ollama run qwen2.5vl:3b-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 Qwen2.5-VL 3B Instruct (4bit, needs ~3.1 GB, Mac only, from [huggingface.co/mlx-community](https://huggingface.co/mlx-community/Qwen2.5-VL-3B-Instruct-4bit/tree/46d4cf06a06ffc1a766c214174f9cbed2f45bcab))
```sh
mlx_lm.chat --model mlx-community/Qwen2.5-VL-3B-Instruct-4bit
```
3. Or start an OpenAI-compatible server at http://localhost:8080/v1
```sh
mlx_lm.server --model mlx-community/Qwen2.5-VL-3B-Instruct-4bit
```
Other versions: 8bit (~4.1 GB, Mac only) `mlx_lm.chat --model mlx-community/Qwen2.5-VL-3B-Instruct-8bit`
Model repository and documentation: [huggingface.co/Qwen](https://huggingface.co/Qwen/Qwen2.5-VL-3B-Instruct)
[interactive version](https://whatmodel.sh/model/qwen2-5-vl-3b) · [JSON](https://whatmodel.sh/api?format=json&all=1)