whatmodel.sh / Qwen3-VL 4B Instruct

Qwen3-VL 4B Instruct

+ contribute

Qwen3-VL 4B Instruct is an open-weights language model from Alibaba, released October 2025. Its published size is 4.44B parameters and a 256k-token context window. Licence: apache-2.0.

Catalogue identity notes

Maker repo Qwen/Qwen3-VL-4B-Instruct @ebb281e (licence tag apache-2.0, created 2025-10-11); the separate Instruct checkpoint, not Qwen3-VL-4B-Thinking. Scored from the Qwen3-VL Technical Report (arXiv 2511.21631v2, Tables 3/4 and 7-10: column 'Qwen3-VL 4B instruct'), whose instruct and thinking columns are the two published checkpoints. Released 2025-10: month the maker's Hugging Face repo was created.

Model information sources

Release date ↗ · Weights licence ↗

4,437,815,808 parameters in the Hugging Face safetensors metadata of Qwen/Qwen3-VL-4B-Instruct @ebb281e (vision encoder included); dense, so active = total.

52.7±0.7#307 of 719 LLMsfrom 27 benchmark results across 3 sources

$ how it compares

The LLMs ranked either side of Qwen3-VL 4B Instruct, out of 828 with a score, then more from Alibaba.

#	model	score	where	RAM	released
366	Gemma 4 E4B Google	53±1	local	~6.9GB	2026-04
367	Ministral 3 8B Mistral	53±1	api $0.15/M +local	~6.5GB	2025-12
368	Qwen3-VL 4B Instruct Alibaba	53±1	local	~4.1GB	2025-10
369	Claude 3 Opus Anthropic	53±0	comparison only		2024-02
370	LFM2.5 2.6B Liquid AI	53±1	local	~3.2GB	2026-07
	more from Alibaba
18	Qwen3.8 Max (2.4T-A95B) Alibaba	78±1	api $3.00/M +local	~380GB	2026-08
34	Qwen3.8 Max (0902) Alibaba	76±1	api $3.00/M		2026-09
37	Qwen3.8-Flash-Next Alibaba	75±1	local	~59GB	2026-08

see all LLMs · find the ones that fit your machine

$ what it's good at

agent tasks▁▁▃▆█▆▅▅▅▄▃▁ #199 of 300 * · 2.0% of its score · 1 test

Beats o1 (2024-12-17), o3-mini, DeepSeek R1 (Jan '25) here, though they rank higher overall. Trails SmolLM3 3B, Jamba 1.5 Mini, Mellum2 12B-A2.5B Instruct here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
BFCL v3 lab	63.3% ✎	█▄··▄█·█▄█▄█ beats 5/13	medium · avg 59%	▇ 2.0%	blunt · at its level

coding▁▁▁▃▆██▇▆▄▃▁ #332 of 458 * · 4.1% of its score · 1 test

Beats GLM-4.7-Flash here, though they rank higher overall. Trails Phi-3-mini-128k-instruct (Apr 2024), DeepSeek LLM 67B Chat, WizardLM 70B V1.0 here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
LiveCodeBench v6 (window unspecified) lab	37.9% ✎	▃▅█▅▅▃▅···▃█ beats 4/16	medium · avg 35%	▇▇ 4.1%	at its level

chat & writing▁▁▁▃▅▄▅█▆▆▇▃ #204 of 393 * · 3.0% of its score · 1 test

Beats Granite 4.2 8B, Qwen3-VL 30B-A3B Thinking, Mistral Small 4 (119B-A6.5B) here, though they rank higher overall. Trails LFM2.5 1.2B Instruct, Gemma 3n E4B, LFM2.5 1.2B Thinking here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
IFEval (aggregation unspecified) lab	82.3% ✎	▃··▃···██▅▃▃ beats 3/11	easy · avg 77%	▇ 3.0%	typical

math▁▂▃▅▆▅▄█▄▁▁▁ #206 of 398 * · 11% of its score · 2 tests

Beats DeepSeek V3 (Dec '24), GPT-4o (2024-08-06), Command A (03-2025) here, though they rank higher overall. Trails Olmo 3 7B Instruct, DeepSeek R1 Distill Llama 8B, Phi-4-mini-reasoning 3.8B here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
MathVista (testmini) lab	73.7% ✎	▃··▃▅·▃█▅▅▅█ beats 7/16	medium · avg 69%	▇▇ 5.8%	at its level
AIME 2025 lab	46.6% ✎	▃·▃▃█··▃▅▃·█ beats 4/12	medium · avg 45%	▇▇ 4.8%	sharp · at its level

vision & documents▁▁▃▂▁▃▄▂▆▇█▄ #113 of 147 · 50% of its score · 15 tests

Beats DeepSeek V4 Flash Vision (Exp), GPT-4 Turbo (2024-04-09), Qwen2-VL 72B Instruct here, though they rank higher overall.

test	result	vs the others in its run	difficulty	counts	why
CharXiv (reasoning questions) lab	39.7% ✎	▄▂▄▆▄▄▆··▄█▂ beats 3/21	medium · avg 39%	▇▇▇▇ 8.8%	at its level
MMMU (val) lab	67.4% ✎	▂···▆▂▂▂▄▄▄█ beats 7/16	medium · avg 61%	▇▇▇ 7.8%	at its level
CharXiv descriptive questions lab	76.2% ✎	▄··▂▂▃·▂▄▄▂█ beats 4/14	easy · avg 74%	▇▇▇ 7.1%	typical
MathVista (testmini) lab	73.7% ✎	▃··▃▅·▃█▅▅▅█ beats 7/16	medium · avg 69%	▇▇ 5.8%	at its level
MMStar lab	69.8% ✎	▂··▄▂▂·▄█▂▄▆ beats 7/16	medium · avg 65%	▇▇ 5.1%	at its level
MMMU Pro lab	53.2% ✎	▃·█·▅▃▃▃▅▅▃█ beats 8/16	medium · avg 46%	▇▇ 4.2%	at its level
ChartArena chart parsing (English, mAP_high average)	33.2% ✎	▁·▁·▃▃▁▇▃█·▄ beats 7/22	hard · avg 27%	▇ 2.3%	blunt · at its level
HallusionBench lab	57.6% ✎	▃·▃·▅▃▅▅▃▅█▅ beats 7/16	medium · avg 55%	▇ 2.0%	blunt · at its level

7 more tests

test	result	vs the others in its run	difficulty	counts	why
ChartQA lab	84.6% ✎	▂··▂·▂▃·▅▂▆█ beats 4/17	easy · avg 85%	▇ 1.6%	blunt
DocVQA lab	95.3% ✎	▃··▅▃▅▃█▃█▅▅ beats 8/17	very easy · avg 94%	▇ 1.5%	blunt · maxed out
RealWorldQA lab	70.9% ✎	▅▃·▃▅▅·█·▃▅█ beats 6/16	easy · avg 70%	▇ 1.5%	blunt · at its level
ScreenSpot-Pro lab	59.5% ✎	▂····▂▄··▂█▆ beats 8/11	medium · avg 46%	▇ 1.2%	blunt · at its level
CC-OCR lab	76.2% ✎	▃··▃▅·▅▅·█▅▃ beats 4/13	easy · avg 74%	▇ 1.0%	blunt
OCRBench lab	88.1% ✎	▅█·▃▃█▃▅▅▃·▃ beats 11/16	easy · avg 84%	▇ 0.5%	blunt
PureDocBench document parsing (Avg3 over Clean/Digital/Real)	67.5% ✎	▂·▁·▂▂▅█·▁▂▂ beats 9/22	medium · avg 62%	▇ 0.1%	blunt · at its level

other tests8 not in a task · 35% of its score

test	result	vs the others in its run	difficulty	counts	why
GPQA (subset unspecified) lab	55.9% ✎	▂····▂·▂▃█▆█ beats 3/14	hard · avg 49%	▇▇▇ 8.1%	sharp · at its level
MMLU-Pro lab	67.1% ✎	▄·▄·▄▄▄▄█▄██ beats 3/12	medium · avg 63%	▇▇▇ 7.4%	at its level
MMLU-Redux lab	81.5% ✎	▃··▃·▃·█▃▅▃▅ beats 2/11	easy · avg 79%	▇▇ 6.2%	typical
SuperGPQA lab	40.3% ✎	▄··▄·▄▄▄█▄██ beats 2/11	medium · avg 37%	▇▇ 4.1%	at its level
AndroidWorld lab	45.3% ✎	▃··▃▅▃▃▃▃▃·█ beats 1/11	medium · avg 43%	▇ 3.0%	blunt · at its level
LVBench lab	56.2% ✎	▅··▃·▅▃▃·█·▃ beats 4/10	hard · avg 51%	▇ 2.8%	at its level
Video-MME (without subtitles) lab	69.3% ✎	▃··█··▅█▃▃▅▃ beats 5/13	medium · avg 66%	▇ 2.8%	blunt · at its level
MVBench lab	68.9% ✎	▂▂▂·▂█··▆▄·▂ beats 5/13	medium · avg 67%	▇ 1.1%	blunt · at its level

$ where to run it

llama.cpp · recommended

version

Install llama.cpp · Homebrew and winget installs name the commands llama-cli and llama-server instead.
macOS / Linux
curl -LsSf https://llama.app/install.sh | sh
Windows
irm https://llama.app/install.ps1 | iex
other ways to install ↗
Run Qwen3-VL 4B Instruct · needs ~2.5 GB · from huggingface.co/Qwen (official)
chat in the terminal
llama cli -hf Qwen/Qwen3-VL-4B-Instruct-GGUF:Q4_K_M
Or start a local server and open the address it prints (http://localhost:9931)
llama serve -hf Qwen/Qwen3-VL-4B-Instruct-GGUF:Q4_K_M

LM Studio

version

Install LM Studio · Open the app once before using the lms command.
download it (macOS / Windows / Linux) ↗
Run Qwen3-VL 4B Instruct · needs ~2.5 GB · from huggingface.co/Qwen (official)
open in LM Studio ↗
Or download it in the terminal
lms get https://huggingface.co/Qwen/Qwen3-VL-4B-Instruct-GGUF@Q4_K_M -y
Then run lms ls to see its name, and lms chat followed by that name

Ollama

version

Install Ollama
download it (macOS / Windows) ↗
Linux
curl -fsSL https://ollama.com/install.sh | sh
Run Qwen3-VL 4B Instruct · needs ~3.3 GB · from ollama.com
chat in the terminal
ollama run qwen3-vl:4b-instruct

MLX (Mac)

version

Install MLX (Mac) · Apple silicon Macs only, text only.
with uv
uv tool install mlx-lm
or
pip install mlx-lm
Run Qwen3-VL 4B Instruct · needs ~3.1 GB, Mac only · from huggingface.co/lmstudio-community
chat in the terminal
mlx_lm.chat --model lmstudio-community/Qwen3-VL-4B-Instruct-MLX-4bit
Or start an OpenAI-compatible server at http://localhost:8080/v1
mlx_lm.server --model lmstudio-community/Qwen3-VL-4B-Instruct-MLX-4bit

Official weights and documentation: huggingface.co/Qwen ↗

$ how the score was worked out · Qwen (Alibaba) vendor-reporte… 97%, ChartArena: Benchmarking Char… 2%, How Far Is Document Parsing f… 0%

Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works

source	results	share of the score	licence
Qwen (Alibaba) vendor-reported benchmark scores for its own models (qwen.ai launch posts, Qwen2.5-VL and Qwen3-VL technical reports, Qwen3.8-Flash-Next card) · lab-reported	25	▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 97%	Factual vendor-reported score (not licensed content)
ChartArena: Benchmarking Chart Parsing across Languages, Scenarios, and Formats (arXiv:2606.01348v3)	1	▇ 2%	CC BY 4.0
How Far Is Document Parsing from Solved? PureDocBench: A Source-Traceable Benchmark across Clean, Degraded, and Real-World Settings (arXiv:2605.07492v2)	1	▇ 0%	CC BY 4.0
prior · pull toward the middle		▇ 0%
weighted average		= 52.7

$ improve this page

Update model information Add a place to run it Add a test result Propose a benchmark Correct a benchmark score Edit data on GitHub

Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.

data updated 2026-10-11 02:20 UTC · sources and licences · json