whatmodel.sh / Qwen2.5-VL 72B Instruct

Qwen2.5-VL 72B Instruct

+ contribute
Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

Qwen2.5-VL 72B Instruct is an open-weights language model from Alibaba, released January 2025. Its published size is 73.4B parameters and a 33k-token context window. Licence: qwen.

Catalogue identity notes

Arena entry 'qwen2.5-vl-72b-instruct' (vision). LMArena's leaderboard table (arena-leaderboard space @6f2eeeb) labels 'qwen2.5-vl-72b-instruct' as 'Qwen2.5-VL-72B-Instruct' (Alibaba, Qwen), linking https://huggingface.co/Qwen/Qwen2.5-VL-72B-Instruct. Maker repo Qwen/Qwen2.5-VL-72B-Instruct @89c8620: licence tag other (qwen). 73,410,777,344 parameters from the Hugging Face safetensors metadata (dense, so active = total). Released 2025-01: month the maker's Hugging Face repo was created (createdAt 2025-01-27).

55.9±0.9#247 of 719 LLMsfrom 29 benchmark results across 9 sources

$ how it compares

The LLMs ranked either side of Qwen2.5-VL 72B Instruct, out of 828 with a score, then more from Alibaba.

#modelscorewhereRAMreleased
296EXAONE Deep 32B LG AI Research56±2local~22GB2025-03
297DeepSeek R1 Distill Qwen 32B DeepSeek56±1api $1.59/M +local~21GB2025-01
298Qwen2.5-VL 72B Instruct Alibaba56±1local~46GB2025-01
299Granite 4.1 30B IBM56±1local~19GB2026-04
300Mistral Small 3.2 24B Mistral56±1api $0.11/M +local~16GB2025-06
more from Alibaba
18Qwen3.8 Max (2.4T-A95B) Alibaba78±1api $3.00/M +local~380GB2026-08
34Qwen3.8 Max (0902) Alibaba76±1api $3.00/M2026-09
37Qwen3.8-Flash-Next Alibaba75±1local~59GB2026-08

see all LLMs · find the ones that fit your machine

$ what it's good at

math▁▂▃▅▆▅▄█▄▁▁▁ #192 of 398 * · 7.8% of its score · 1 test

Beats MiniMax-M2.1, Qwen3 235B-A22B (Apr 2025, hybrid), DeepSeek V3 (Dec '24) here, though they rank higher overall. Trails Phi-4-mini-reasoning 3.8B, DeepSeek R1 Distill Qwen 7B, OpenThinker3 7B here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
MathVista (testmini) lab74.8% ✎▃··▃▅·▃█▅▅▅█ beats 8/16medium · avg 69%▇▇▇ 7.8%at its level
vision & documents▁▁▃▂▁▃▄▂▆▇█▄ #103 of 147 · 92% of its score · 20 tests

Beats DeepSeek V4 Flash Vision (Exp), GPT-4o (2024-08-06) here, though they rank higher overall. Trails Llama 4 Scout (109B-A17B), Mistral Small 3.2 24B here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
CharXiv (reasoning questions) lab49.7% ✎▄▂▄▆▄▄▆··▄█▂ beats 7/21medium · avg 39%▇▇▇▇▇ 13%at its level
MMMU (val) lab70.2% ✎▂···▆▂▂▂▄▄▄█ beats 9/16medium · avg 61%▇▇▇▇ 11%at its level
MMStar lab +170.8% ✎▂··▄▂▂·▄█▂▄▆ beats 8/16medium · avg 65%▇▇▇▇ 9.4%at its level · several runs
MMMU Pro lab +151.1% ✎▃·█·▅▃▃▃▅▅▃█ beats 7/16medium · avg 46%▇▇▇▇ 9.2%at its level · several runs
CharXiv descriptive questions lab87.4% ✎▄··▂▂▂·▂▄▄▃█ beats 10/14easy · avg 74%▇▇▇▇ 9.1%typical
MathVista (testmini) lab74.8% ✎▃··▃▅·▃█▅▅▅█ beats 8/16medium · avg 69%▇▇▇ 7.8%at its level
Arena vision (style controlled)1121.2 ✎▁▁▂▂▂▂▄▅▅▆██ beats 36/145avg 1068▇▇ 5.0%blunt
ChartArena chart parsing (English, mAP_high average)36.6% ✎▁·▁·▃▃▁▇▃█·▄ beats 10/22hard · avg 27%▇ 3.6%blunt · at its level
12 more tests
testresultvs the others in its rundifficultycountswhy
CapArena-Auto detailed image captioning (Score_Avg, GPT-4o judge)35.3300 ✎▃▃▃·▃█·▅▃▃▃▃ beats 10/12avg -0.671▇ 3.4%typical
HallusionBench lab55.2% ✎▃·▃·▅▃▅▅▃▅█▅ beats 6/16medium · avg 55%▇ 3.0%blunt · at its level
Kaleidoscope (in-language multimodal exams, 18 languages, overall accuracy)52.9% ✎█·█···█·█·██ beats 3/5medium · avg 47%▇ 2.9%at its level
DocVQA lab +196.4% ✎▃··▅▃▅▃█▃█▅▅ beats 13/17very easy · avg 94%▇ 2.8%blunt · maxed out · several runs
ChartQA lab89.5% ✎▂··▂·▂▃·▅▂▆█ beats 14/17easy · avg 85%▇ 2.2%blunt
RealWorldQA lab75.7% ✎▅▃·▃▅▅·█·▃▅█ beats 11/16easy · avg 70%▇ 2.1%blunt · at its level
CC-OCR lab +179.8% ✎▃··▃▅·▅▅·█▅▃ beats 9/13easy · avg 74%▇ 1.9%blunt · several runs
VMMU (Vietnamese multimodal understanding and reasoning, mean accuracy over 7 tasks)51.7% ✎▅······▃·▅·█ beats 6/7very hard · avg 36%▇ 1.9%noisy
MSEarth-Bench Earth-science figure MCQ (Overall accuracy)50.6% ✎█·██····██·█ beats 1/5hard · avg 48%▇ 1.2%blunt · at its level
ScreenSpot-Pro lab43.6% ✎▂▃▅▆██▅▂···▂ beats 2/24medium · avg 44%▇ 1.0%blunt · at its level · small run
OCRBench lab88.5% ✎▅█·▃▃█▃▅▅▃·▃ beats 12/16easy · avg 84%▇ 0.7%blunt
TextVQA lab83.5% ✎█······▄·▄▄▄ beats 2/5easy · avg 80%▇ 0.4%blunt
other tests5 not in a task · 7.5% of its score
testresultvs the others in its rundifficultycountswhy
Video-MME (with subtitles) lab79.1% ✎▁▁▁▁▁·▁▃▃▁▁█ beats 8/16medium · avg 73%▇ 1.9%at its level · small run
LVBench lab47.3% ✎▇·█▅▅▄▄··▃▁▁ beats 1/28hard · avg 49%▇ 1.6%at its level · small run
Video-MME (without subtitles) lab73.3% ✎▁▃▁▃▁·█▄▄▅·▇ beats 15/27medium · avg 67%▇ 1.4%blunt · at its level · small run
Open PT-LLM Leaderboard average (9 Portuguese tasks)79.1% ✎▃▁▁▁▁▁▁▁▂█▅▁ beats 251/276medium · avg 73%▇ 1.4%at its level · noisy
AndroidWorld lab35.0% ✎▁·▃▁▂▃█▅▂▂·▂ beats 1/24hard · avg 34%▇ 1.3%blunt · at its level · small run

$ where to run it

llama.cpp · recommended

version

  1. 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 ↗

  2. Run Qwen2.5-VL 72B Instruct · needs ~47.4 GB · from huggingface.co/lmstudio-community
    chat in the terminal
    llama cli -hf lmstudio-community/Qwen2.5-VL-72B-Instruct-GGUF:Q4_K_M
  3. Or start a local server and open the address it prints (http://localhost:9931)
    llama serve -hf lmstudio-community/Qwen2.5-VL-72B-Instruct-GGUF:Q4_K_M

LM Studio

version

  1. Install LM Studio · Open the app once before using the lms command.

    download it (macOS / Windows / Linux) ↗

  2. Run Qwen2.5-VL 72B Instruct · needs ~47.4 GB · from huggingface.co/lmstudio-community

    open in LM Studio ↗

  3. Or download it in the terminal
    lms get https://huggingface.co/lmstudio-community/Qwen2.5-VL-72B-Instruct-GGUF@Q4_K_M -y
  4. Then run lms ls to see its name, and lms chat followed by that name

Ollama

version

  1. Install Ollama

    download it (macOS / Windows) ↗

    Linux
    curl -fsSL https://ollama.com/install.sh | sh
  2. Run Qwen2.5-VL 72B Instruct · needs ~44.54 GB · from ollama.com
    chat in the terminal
    ollama run qwen2.5vl:72b-q4_K_M

MLX (Mac)

version

  1. Install MLX (Mac) · Apple silicon Macs only, text only.
    with uv
    uv tool install mlx-lm
    or
    pip install mlx-lm
  2. Run Qwen2.5-VL 72B Instruct · needs ~42.3 GB, Mac only · from huggingface.co/mlx-community
    chat in the terminal
    mlx_lm.chat --model mlx-community/Qwen2.5-VL-72B-Instruct-4bit
  3. Or start an OpenAI-compatible server at http://localhost:8080/v1
    mlx_lm.server --model mlx-community/Qwen2.5-VL-72B-Instruct-4bit

Official weights and documentation: huggingface.co/Qwen ↗

$ how the score was worked out · Qwen (Alibaba) vendor-reporte… 67%, Qwen2.5-VL-32B-Instruct model… 13%, LMArena leaderboard dataset 5% and 6 more

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

sourceresultsshare of the scorelicence
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-reported13▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 67%Factual vendor-reported score (not licensed content)
Qwen2.5-VL-32B-Instruct model card · lab-reported9▇▇▇▇ 13%Apache-2.0
LMArena leaderboard dataset1▇ 5%CC BY 4.0
ChartArena: Benchmarking Chart Parsing across Languages, Scenarios, and Formats (arXiv:2606.01348v3)1▇ 4%CC BY 4.0
CapArena: Benchmarking and Analyzing Detailed Image Captioning in the LLM Era (arXiv:2503.12329v1)1▇ 3%CC-BY-4.0
Kaleidoscope: In-language Exams for Massively Multilingual Vision Evaluation (arXiv:2504.07072v2)1▇ 3%CC-BY-4.0
VMMU: A Vietnamese Multitask Multimodal Understanding and Reasoning Benchmark (arXiv:2508.13680v5)1▇ 2%CC-BY-4.0
Open Portuguese LLM Leaderboard (archived)1▇ 1%Apache-2.0
MSEarth: A Multimodal Benchmark for Earth Science Phenomenon Discovery with MLLMs (arXiv:2505.20740v3)1▇ 1%CC-BY-4.0
prior · pull toward the middle▇ 0%
weighted average= 55.9

$ improve this page

Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit 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 · sources and licences · json