whatmodel.sh / Qwen2-VL 7B Instruct

Qwen2-VL 7B Instruct

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

Qwen2-VL 7B Instruct is an open-weights language model from Alibaba, released August 2024. Its published size is 8.29B parameters. Licence: apache-2.0.

Catalogue identity notes

Comparison-only. Arena entry 'qwen2-vl-7b-instruct' (vision). LMArena's leaderboard table (arena-leaderboard space @6f2eeeb) labels 'qwen2-vl-7b-instruct' as 'Qwen2-VL-7B-Instruct' (Aliaba, Apache 2.0), linking https://huggingface.co/Qwen/Qwen2-VL-7B-Instruct. Maker repo Qwen/Qwen2-VL-7B-Instruct @eed1309: licence tag apache-2.0. 8,291,375,616 parameters from the Hugging Face safetensors metadata (dense, so active = total). Released 2024-08: month the maker's Hugging Face repo was created (createdAt 2024-08-28).

45.9±1.0#425 of 719 LLMsfrom 27 benchmark results across 6 sources

$ how it compares

The LLMs ranked either side of Qwen2-VL 7B Instruct, out of 828 with a score, then more from Alibaba.

#modelscorewhereRAMreleased
504T5 Base Google46±9local~2.4GB2019-10
505OpenChat 3.5 0106 OpenChat46±1local~6.1GB2024-01
506Qwen2-VL 7B Instruct Alibaba46±1local~6.4GB2024-08
507Qwen2.5-Math 1.5B Instruct Alibaba46±2local~2.4GB2024-09
508Yi 1.5 34B Chat 01.AI46±1local~22GB2024-05
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

coding▁▁▁▃▆██▇▆▄▃▁ #320 of 458 * · 2.4% of its score · 1 test

Beats GLM-4.7-Flash, Qwen3-VL 8B Instruct, Qwen3-VL 4B Instruct here, though they rank higher overall. Trails DeepSeek LLM 67B Chat, WizardLM 70B V1.0, Llama 3.2 11B Vision Instruct here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
HumanEval62.2% ✎▁▁▁▁▁▃▅▅█▃▂▁ beats 102/190easy · avg 72%▇ 2.4%sharp · at its level · noisy
math▁▂▃▅▆▅▄█▄▁▁▁ #335 of 398 · 19% of its score · 3 tests

Beats GPT-3.5 Turbo (0125) here, though they rank higher overall. Trails Qwen3 0.6B, Gemma 7B (base), LFM2 700M here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
MathVista (testmini) lab +158.2% ✎▃··▃▅·▃█▅▅▅█ beats 1/16medium · avg 69%▇▇▇▇▇ 13%at its level · several runs
MATH-Vision lab16.3% ✎▁▃·▁··▄▁▃▇▇█ beats 1/24hard · avg 27%▇▇ 4.1%small run
MATH (subset unspecified)27.9% ✎▂▄▄▃█▇▅▆▄▃▄▁ beats 11/80medium · avg 55%▇ 2.6%at its level
vision & documents▁▁▃▃▁▃▄▂▆▇█▄ #128 of 147 · 79% of its score · 17 tests

Beats Nova Lite, Gemma 4 E2B here, though they rank higher overall.

testresultvs the others in its rundifficultycountswhy
MMMU (val) lab +154.1% ✎▂···▆▂▂▂▄▄▄█ beats 3/16medium · avg 61%▇▇▇▇▇▇ 15%at its level · several runs
MathVista (testmini) lab +158.2% ✎▃··▃▅·▃█▅▅▅█ beats 1/16medium · avg 69%▇▇▇▇▇ 13%at its level · several runs
MMStar lab60.7% ✎▂··▄▃▂·▄█▂▄▆ beats 3/16medium · avg 65%▇▇▇▇ 8.8%at its level
MMMU Pro lab43.5% ✎▃·█·▅▃▃▃▅▅▃█ beats 5/16medium · avg 46%▇▇▇ 6.3%at its level
Arena vision (style controlled)1030.8 ✎▁▁▂▂▃▂▄▅▅▆██ beats 18/145avg 1068▇▇ 5.7%blunt
MATH-Vision lab16.3% ✎▁▃·▁··▄▁▃▇▇█ beats 1/24hard · avg 27%▇▇ 4.1%small run
CapArena-Auto detailed image captioning (Score_Avg, GPT-4o judge)-49.0000 ✎▃▃▃·▃█·▅▃▃▃▃ beats 1/12avg -0.671▇▇ 3.8%typical
HallusionBench lab50.6% ✎▃·▃·▅▃▅▅▃▅█▅ beats 2/16medium · avg 55%▇ 3.5%blunt · at its level
9 more tests
testresultvs the others in its rundifficultycountswhy
ChartQA lab +183.0% ✎▂··▂·▃▃·▅▂▆█ beats 2/17easy · avg 85%▇ 3.4%blunt · several runs
InfographicVQA lab76.5% ✎▃·▃▅█▃▃▃▃▃▃▃ beats 7/13easy · avg 80%▇ 3.2%at its level · small run
DocVQA lab94.5% ✎▃··▅▃▅▃█▃█▅▅ beats 5/17very easy · avg 94%▇ 3.2%blunt · maxed out
MM-MT-Bench (Mistral, GPT-4o judge)5.4500 MM-MT-Bench score (1-10) ✎▅··▃·█·▃▃··▅ beats 3/9avg 5.740 MM-MT-Bench score (1-10)▇ 2.8%sharp · noisy
RealWorldQA lab70.1% ✎▅▃·▃▅▅·█·▃▅█ beats 5/16easy · avg 70%▇ 2.7%blunt · at its level
VQAv2 (val)75.9% ✎▃···█▃··▅▃▃▃ beats 6/9easy · avg 74%▇ 1.7%blunt · at its level
OCRBench lab86.6% ✎▅█·▃▃█▃▅▅▃·▃ beats 9/16easy · avg 84%▇ 1.0%blunt
Ai2 Molmo human preference Elo (Sept 2024, 27 VLMs)1025.0 ✎▄▄·▄·▄▄▄·▄·█ beats 4/8avg 1037▇ 0.9%blunt
TextVQA lab84.3% ✎█······▄·▄▄▄ beats 3/5easy · avg 80%▇ 0.6%blunt
other tests5 not in a task · 15% of its score
testresultvs the others in its rundifficultycountswhy
MMLU68.5% ✎▁▁▁▁▁▂▃▅█▃▆▃ beats 104/183easy · avg 76%▇▇ 5.5%at its level
Video-MME (without subtitles) lab63.3% ✎▃··█··▅█▃▃▅▃ beats 3/13medium · avg 66%▇▇ 4.8%blunt · at its level
Video-MME (with subtitles) lab69.0% ✎▁▁▁▁▃·▁▃▁▁▁█ beats 4/16medium · avg 73%▇ 2.3%at its level · small run
MVBench lab67.0% ✎▂▂▂·▃█··▆▄·▂ beats 3/13medium · avg 67%▇ 1.9%blunt · at its level
Vectara hallucination leaderboard, 2025 dataset (HHEM-2.3, factual consistency rate)95.8% ✎▁▁▂▃▄▄█▆▃▂▂▁ beats 82/140very easy · avg 94%▇ 0.6%blunt · maxed out · noisy

$ 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-VL 7B Instruct · needs ~5.4 GB · from huggingface.co/ggml-org
    chat in the terminal
    llama cli -hf ggml-org/Qwen2-VL-7B-Instruct-GGUF:Q4_K_M
  3. Or start a local server and open the address it prints (http://localhost:9931)
    llama serve -hf ggml-org/Qwen2-VL-7B-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-VL 7B Instruct · needs ~5.4 GB · from huggingface.co/ggml-org

    open in LM Studio ↗

  3. Or download it in the terminal
    lms get https://huggingface.co/ggml-org/Qwen2-VL-7B-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-VL 7B Instruct · needs ~5.4 GB · from huggingface.co/ggml-org
    chat in the terminal
    ollama run hf.co/ggml-org/Qwen2-VL-7B-Instruct-GGUF: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-VL 7B Instruct · needs ~4.7 GB, Mac only · from huggingface.co/mlx-community
    chat in the terminal
    mlx_lm.chat --model mlx-community/Qwen2-VL-7B-Instruct-4bit
  3. Or start an OpenAI-compatible server at http://localhost:8080/v1
    mlx_lm.server --model mlx-community/Qwen2-VL-7B-Instruct-4bit

Official weights and documentation: huggingface.co/Qwen ↗

$ how the score was worked out · Qwen (Alibaba) vendor-reporte… 70%, Pixtral 12B (Mistral AI, arXi… 19%, LMArena leaderboard dataset 6% and 3 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-reported15▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 70%Factual vendor-reported score (not licensed content)
Pixtral 12B (Mistral AI, arXiv 2410.07073v2)8▇▇▇▇▇▇ 19%CC BY 4.0
LMArena leaderboard dataset1▇▇ 6%CC BY 4.0
CapArena: Benchmarking and Analyzing Detailed Image Captioning in the LLM Era (arXiv:2503.12329v1)1▇ 4%CC-BY-4.0
Ai2 Molmo-72B-0924 model card: Human Preference Elo Rating column1▇ 1%Apache-2.0
Vectara hallucination leaderboard (vectara/hallucination-leaderboard repository)1▇ 1%Apache-2.0
prior · pull toward the middle▇ 0%
weighted average= 45.9

$ 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