whatmodel.sh / Qwen3-VL 2B Thinking

Qwen3-VL 2B Thinking

+ contribute

Qwen3-VL 2B Thinking is an open-weights language model from Alibaba, released October 2025. Its published size is 2.13B parameters and a 256k-token context window. Licence: apache-2.0.

Catalogue identity notes

Maker repo Qwen/Qwen3-VL-2B-Thinking @33e0ad9 (licence tag apache-2.0, created 2025-10-19); the separate Thinking checkpoint, not Qwen3-VL-2B-Instruct. Scored from the Qwen3-VL Technical Report (arXiv 2511.21631v2, Tables 3/4 and 7-10: column 'Qwen3-VL 2B thinking'), 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 ↗

2,127,532,032 parameters in the Hugging Face safetensors metadata of Qwen/Qwen3-VL-2B-Thinking @33e0ad9 (vision encoder included); dense, so active = total.

48.9±0.7#363 of 719 LLMsfrom 26 benchmark results across 2 sources

$ how it compares

The LLMs ranked either side of Qwen3-VL 2B Thinking, out of 828 with a score, then more from Alibaba.

#	model	score	where	RAM	released
432	Artigenz-Coder-DS-6.7B Artigenz	49±1	local	~5.8GB	2024-04
433	LFM2.5 1.2B Thinking Liquid AI	49±2	local	~2.2GB	2026-01
434	Qwen3-VL 2B Thinking Alibaba	49±1	local	~2.7GB	2025-10
435	speechless-codellama-34b-v2.0 uukuguy	49±1	local	~23GB	2023-10
436	Pixtral 12B (2409) Mistral	49±2	local	~9.6GB	2024-09
	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▁▁▃▆█▆▅▅▅▄▃▁ #228 of 300 * · 2.0% of its score · 1 test

Beats o3-mini, Nova 2 Lite, Ministral 3 14B Reasoning 2512 here, though they rank higher overall. Trails SmolLM3 3B, Phi-3.5-mini-instruct, Jamba 1.5 Mini here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
BFCL v3 lab	57.2% ✎	█▄··▄█·█▄█▄█ beats 2/13	medium · avg 59%	▇ 2.0%	blunt · at its level

coding▁▁▁▃▆██▇▆▄▃▁ #369 of 458 * · 3.7% of its score · 1 test

Beats LFM2.5 8B-A1B here, though they rank higher overall. Trails Qwen2.5 1.5B Instruct, WizardLM 13B V1.2, Granite 3.0 2B Instruct here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
LiveCodeBench v6 (window unspecified) lab	29.3% ✎	▃▅█▅▅▃▅···▃█ beats 1/16	medium · avg 35%	▇ 3.7%	at its level

chat & writing▁▁▁▃▅▄▅█▆▆▇▃ #252 of 393 * · 3.3% of its score · 1 test

Beats Mistral Small 4 (119B-A6.5B), Ministral 3 14B Reasoning 2512, Qwen2.5 Coder 32B Instruct here, though they rank higher overall. Trails Flan-T5 XXL, Granite 4.0 H 1B, Granite 4.0 1B here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
IFEval (aggregation unspecified) lab	75.1% ✎	▃··▃···██▅▃▃ beats 1/11	easy · avg 77%	▇ 3.3%	typical

math▁▂▃▅▆▅▄█▄▁▁▁ #220 of 398 * · 11% of its score · 2 tests

Beats GPT-4o (2024-08-06), Command A (03-2025), GPT-4o (2024-11-20) 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	73.6% ✎	▃··▃▅·▃█▅▅▅█ beats 6/16	medium · avg 69%	▇▇ 6.1%	at its level
AIME 2025 lab	39.0% ✎	▃·▃▃█··▃▅▃·█ beats 2/12	medium · avg 45%	▇▇ 4.6%	sharp · at its level

vision & documents▁▁▃▂▃▃▄▂▆▇█▄ #120 of 147 · 51% of its score · 14 tests

Beats DeepSeek V4 Flash Vision (Exp), Gemma 3 12B, Llama 3.2 90B Vision Instruct here, though they rank higher overall. Trails Qwen3.5 2B here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
CharXiv (reasoning questions) lab	37.1% ✎	▄▃▄▆▄▄▆··▄█▂ beats 2/21	medium · avg 39%	▇▇▇ 8.2%	at its level
CharXiv descriptive questions lab	70.1% ✎	▄··▃▂▂·▂▄▄▂█ beats 2/14	easy · avg 74%	▇▇▇ 8.1%	at its level
MMMU (val) lab	61.4% ✎	▂···▆▂▃▂▄▄▄█ beats 5/16	medium · avg 61%	▇▇▇ 7.8%	at its level
MathVista (testmini) lab	73.6% ✎	▃··▃▅·▃█▅▅▅█ beats 6/16	medium · avg 69%	▇▇ 6.1%	at its level
MMStar lab	68.1% ✎	▂··▄▂▂·▄█▂▄▆ beats 5/16	medium · avg 65%	▇▇ 5.1%	at its level
MMMU Pro lab	42.5% ✎	▃·█·▅▃▃▃▅▅▃█ beats 4/16	medium · avg 46%	▇▇ 3.9%	at its level
HallusionBench lab	54.9% ✎	▃·▃·▅▃▅▅▃▅█▅ beats 5/16	medium · avg 55%	▇ 2.0%	blunt · at its level
ChartArena chart parsing (English, mAP_high average)	26.5% ✎	▁·▁·▃▃▁▇▃█·▄ beats 4/22	hard · avg 27%	▇ 2.0%	blunt · at its level

6 more tests

test	result	vs the others in its run	difficulty	counts	why
ChartQA lab	86.6% ✎	▂··▂·▂▃·▅▂▆█ beats 5/17	easy · avg 85%	▇ 1.8%	blunt
DocVQA lab	92.9% ✎	▃··▅▃▅▃█▃█▅▅ beats 1/17	very easy · avg 94%	▇ 1.7%	blunt · maxed out
RealWorldQA lab	69.5% ✎	▅▃·▃▅▅·█·▃▅█ beats 4/16	easy · avg 70%	▇ 1.5%	blunt · at its level
ScreenSpot-Pro lab	32.2% ✎	▃····▂▄··▂█▆ beats 0/11	medium · avg 46%	▇ 1.2%	blunt · at its level
CC-OCR lab	68.3% ✎	▃··▃▅·▅▅·█▅▃ beats 0/13	easy · avg 74%	▇ 1.0%	blunt · at its level
OCRBench lab	79.2% ✎	▅█·▃▃█▃▅▅▃·▃ beats 0/16	easy · avg 84%	▇ 0.6%	blunt

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

test	result	vs the others in its run	difficulty	counts	why
MMLU-Pro lab	62.3% ✎	▄·▄·▄▄▄▄█▄██ beats 2/12	medium · avg 63%	▇▇▇ 7.7%	at its level
GPQA (subset unspecified) lab	49.5% ✎	▂····▂·▃▂█▆█ beats 2/14	hard · avg 49%	▇▇▇ 7.0%	sharp · at its level
MMLU-Redux lab	76.9% ✎	▃··▃·▃·█▃▅▃▅ beats 1/11	easy · avg 79%	▇▇▇ 7.0%	at its level
SuperGPQA lab	34.6% ✎	▄··▄·▄▄▄█▄██ beats 1/11	medium · avg 37%	▇▇ 3.8%	at its level
AndroidWorld lab	46.1% ✎	▃··▃▅▃▃▃▃▃·█ beats 2/11	medium · avg 43%	▇ 2.9%	blunt · at its level
Video-MME (without subtitles) lab	62.1% ✎	▃··█··▅█▃▃▅▃ beats 2/13	medium · avg 66%	▇ 2.8%	blunt · at its level
LVBench lab	47.6% ✎	▅··▃·▅▃▃·█·▃ beats 1/10	hard · avg 51%	▇ 2.6%	at its level
MVBench lab	64.5% ✎	▂▂▃·▂█··▆▄·▂ beats 2/13	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 2B Thinking · needs ~1.1 GB · from huggingface.co/Qwen (official)
chat in the terminal
llama cli -hf Qwen/Qwen3-VL-2B-Thinking-GGUF:Q4_K_M
Or start a local server and open the address it prints (http://localhost:9931)
llama serve -hf Qwen/Qwen3-VL-2B-Thinking-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 2B Thinking · needs ~1.1 GB · from huggingface.co/Qwen (official)
open in LM Studio ↗
Or download it in the terminal
lms get https://huggingface.co/Qwen/Qwen3-VL-2B-Thinking-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 2B Thinking · needs ~1.9 GB · from ollama.com
chat in the terminal
ollama run qwen3-vl:2b-thinking

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 2B Thinking · needs ~1.8 GB, Mac only · from huggingface.co/mlx-community
chat in the terminal
mlx_lm.chat --model mlx-community/Qwen3-VL-2B-Thinking-4bit
Or start an OpenAI-compatible server at http://localhost:8080/v1
mlx_lm.server --model mlx-community/Qwen3-VL-2B-Thinking-4bit

Official weights and documentation: huggingface.co/Qwen ↗

$ how the score was worked out · Qwen (Alibaba) vendor-reporte… 98%, ChartArena: Benchmarking Char… 2%

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	▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 98%	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
prior · pull toward the middle		▇ 0%
weighted average		= 48.9

$ 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