[whatmodel.sh](../../README.md) / Qwen3-VL 32B Instruct # Qwen3-VL 32B Instruct Alibaba · released 2025-10 · 33.4B params · apache-2.0 **score 59.8** ±0.7 · #207 of 732 · from 26 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 Qwen3-VL 32B Instruct among every model on that task or test, worst to best.}

task                vs the others          rank  counts  tests
agent tasks         ▁▁▃▆█●▅▅▄▄▂▁   #172 of 313*    2.1%      1
coding              ▁▁▁▃▅●█▇▅▄▃▁   #311 of 469*    4.3%      1
chat & writing      ▁▁▁▃▆▄▆●▆▇▇▃   #185 of 402*    2.6%      1
math                ▁▁▃▅█▆●█▄▂▁▁   #162 of 431*    9.0%      2
vision & documents  ▁▁▂▂▁▃▄●▆▇█▄     #92 of 153     46%     13

agent tasks · #172 of 313 · 2.1% of its score

Beats [o1 (2024-12-17)](o1-2024-12-17.md), [o3-mini](o3-mini.md), [DeepSeek R1 0528](deepseek-r1-0528.md) here, though they rank higher overall. Trails [SmolLM3 3B](smollm3-3b.md), [Jamba 1.5 Mini](jamba-1-5-mini.md), [Jamba 1.5 Large](jamba-1-5-large.md) here, though they rank lower overall.

test             result  vs the others in its run  difficulty        counts  why
BFCL v3 (maker)   70.2%  █▄··▄█·█▄█●█ beats 11/13  medium · avg 59%    2.1%  blunt · at its level

coding · #311 of 469 · 4.3% of its score

Trails [DeepSeek LLM 67B Chat](deepseek-llm-67b-chat.md), [WizardLM 70B V1.0](wizardlm-70b-v1-0.md), [Yi 1.5 9B Chat](yi-1-5-9b-chat.md) here, though they rank lower overall.

test                                result  vs the others in its run  difficulty      counts  why
LiveCodeBench v6 (window unspecif…   43.8%  ▃▅█●▅▃▅···▃█ beats 7/16   hard · avg 35%    4.3%  at its level

chat & writing · #185 of 402 · 2.6% of its score

Beats [Seed-OSS 36B Instruct](seed-oss-36b.md), [Gemma 2 9B IT SimPO](gemma-2-9b-simpo.md), [Nemotron-Cascade-14B-Thinking](nemotron-cascade-14b-thinking.md) here, though they rank higher overall. Trails [LFM2.5 1.2B Instruct](lfm2-5-1-2b-instruct.md), [LFM2.5 1.2B Thinking](lfm2-5-1-2b-thinking.md), [LFM2.5 8B-A1B](lfm2-5-8b-a1b.md) here, though they rank lower overall.

test                                result  vs the others in its run  difficulty      counts  why
IFEval (aggregation unspecified) …   84.7%  ▃··▃···██●▃▃ beats 8/11   easy · avg 77%    2.6%  typical

math · #162 of 431 · 9.0% of its score

Beats [MiMo-V2.5 (310B-A15B)](mimo-v2-5.md), [MiniMax-M2.7](minimax-m2-7.md), [MiniMax-M2.1](minimax-m2-1.md) here, though they rank higher overall. Trails [Olmo 3 7B Think](olmo-3-7b-think.md), [Qwen3-VL 4B Thinking](qwen3-vl-4b-thinking.md), [Nemotron 3 Nano 30B-A3B](nemotron-3-nano-30b-a3b.md) here, though they rank lower overall.

test                          result  vs the others in its run  difficulty        counts  why
MathVista (testmini) (maker)   83.8%  ▃··▃▅·▃█▅▅▅● beats 14/16  medium · avg 69%    4.8%  typical
AIME 2025 (maker)              66.2%  ▃·▃▃█··●▅▃·█ beats 6/12   medium · avg 44%    4.2%  sharp · at its level

vision & documents · #92 of 153 · 46% of its score

Beats [DeepSeek V4 Flash Vision (Exp)](deepseek-v4-flash-vision-exp.md), [GLM-4.6V](glm-4-6v.md), [GPT-5 nano](gpt-5-nano.md) here, though they rank higher overall.

test                                result  vs the others in its run  difficulty           counts  why
CharXiv (reasoning questions) (ma…   62.8%  ▄▂▄▆▄▄●··▄█▂ beats 12/21  medium · avg 38%        10%  at its level
MMMU (val) (maker)                   76.0%  ▂···▆▂▂▂▄▄▄● beats 13/16  medium · avg 61%       7.4%  at its level
CharXiv descriptive questions (ma…   90.5%  ▄··▂▂▂·▂▄▄▂● beats 13/14  easy · avg 74%         5.4%  typical
MMStar (maker)                       77.7%  ▂··▄▂▂·▄█▂▄● beats 14/16  medium · avg 65%       5.0%  at its level
MathVista (testmini) (maker)         83.8%  ▃··▃▅·▃█▅▅▅● beats 14/16  medium · avg 69%       4.8%  typical
MMMU Pro (maker)                     65.3%  ▃·█·▅▃▃▃▅▅▃● beats 14/16  medium · avg 46%       4.6%  at its level
HallusionBench (maker)               63.8%  ▃·▃·▅▃▅▅▃▅●▅ beats 12/16  medium · avg 54%       2.1%  blunt · at its level
ChartQA (maker)                      88.5%  ▂··▂·▂▃·▅▂●█ beats 9/17   easy · avg 85%         1.5%  blunt
RealWorldQA (maker)                  79.0%  ▅▃·▃▅▅·█·▃▅● beats 15/16  easy · avg 70%         1.5%  blunt
ScreenSpot-Pro (maker)               57.9%  ▂····▂▄··▂●▆ beats 7/11   medium · avg 46%       1.3%  blunt · at its level
DocVQA (maker)                       96.9%  ▃··▅▃▅▃█▃█▅● beats 16/17  very easy · avg 94%    1.3%  blunt · maxed out
CC-OCR (maker)                       80.3%  ▃··▃▅·▅▅·█●▃ beats 11/13  easy · avg 74%         1.0%  blunt
OCRBench (maker)                     89.5%  ▅█·▃▃█▃▅●▃·▃ beats 13/16  easy · avg 84%         0.5%  blunt

other tests (9)

test                                result  vs the others in its run  difficulty        counts  why
GPQA (subset unspecified) (maker)    68.9%  ▂····▂·▂▂█●█ beats 8/14   hard · avg 49%      8.8%  sharp · at its level
MMLU-Pro (maker)                     78.6%  ▄·▄·▄▄▄▄█▄●█ beats 10/12  medium · avg 63%    6.4%  typical
WebCompass (overall)                 35.2%  ▅·●▃▃···▃·▃█ beats 2/9    hard · avg 18%      5.4%  at its level
MMLU-Redux (maker)                   89.8%  ▃··▃·▃·█▃▅●▅ beats 9/11   easy · avg 79%      4.7%  typical
SuperGPQA (maker)                    54.6%  ▄··▄·▄▄▄█▄●█ beats 9/11   medium · avg 37%    4.4%  at its level
AndroidWorld (maker)                 57.3%  ▃··▃▅▃▃▃▃●·█ beats 8/11   medium · avg 43%    3.2%  at its level
LVBench (maker)                      63.8%  ▅··▃·▅▃▃·●·▃ beats 9/10   hard · avg 51%      3.2%  at its level
Video-MME (without subtitles) (ma…   76.6%  ▃··█··▅█▃▃●▃ beats 11/13  medium · avg 66%    2.8%  blunt · at its level
MVBench (maker)                      72.8%  ▂▂▂·▂█··●▄·▂ beats 10/13  medium · avg 67%    1.2%  blunt · at its level

## How the score was worked out Each result implies a score on its own, after adjusting for how hard that test is and who ran it. The model's 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 is added up by who ran the tests. [Method](../method.md)

source                                    results  share of the score
Qwen (submitted / maker-reported)              25   94% ▇▇▇▇▇▇▇▇▇▇▇
WebCompass: Towards Multimodal Web Codi…        1    5% ▇
prior: pull toward the middle                        0% ▇
weighted average                                   = 59.8

_{Licences: WebCompass: Towards Multimodal Web Coding Evaluation for Code Language Models ([CC BY-SA 4.0](https://creativecommons.org/licenses/by-sa/4.0/))} ## How to run it ### Through an API api (Alibaba Cloud Model Studio) · $0.28/M blended · $0.16 in / $0.64 out per 1M tokens · [alibabacloud.com](https://www.alibabacloud.com/help/en/model-studio/model-pricing) _{[price source](https://www.alibabacloud.com/help/en/model-studio/model-pricing): USD per 1M tokens, Alibaba Cloud Model Studio, Singapore (International deployment scope), API model qwen3-vl-32b-instruct (Non-Thinking mode only): input $0.16, output $0.64. Chinese-mainland/Global (US) deployments priced differently and not used, following the International convention of other Alibaba rows; free quota, batch and context-cache discounts excluded. Observed 2026-10-09. · [context source](https://huggingface.co/Qwen/Qwen3-VL-32B-Instruct/blob/0cfaf48183f594c314753d30a4c4974bc75f3ccb/README.md): Official Qwen model card: 'Native 256K context, expandable to 1M'. Observed 2026-10-09.} ### 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 Qwen3-VL 32B Instruct (Q4_K_M, needs ~19.8 GB, from [huggingface.co/Qwen](https://huggingface.co/Qwen/Qwen3-VL-32B-Instruct-GGUF/tree/e3e1fe0c76de7ee58ea65db420c643adfe2e457c)) ```sh llama cli -hf Qwen/Qwen3-VL-32B-Instruct-GGUF:Q4_K_M ``` 3. Or start a local server and open the address it prints (http://localhost:9931) ```sh llama serve -hf Qwen/Qwen3-VL-32B-Instruct-GGUF:Q4_K_M ``` Other versions: Q8_0 (~34.8 GB) `llama cli -hf Qwen/Qwen3-VL-32B-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 Qwen3-VL 32B Instruct (Q4_K_M, needs ~19.8 GB, from [huggingface.co/Qwen](https://huggingface.co/Qwen/Qwen3-VL-32B-Instruct-GGUF/tree/e3e1fe0c76de7ee58ea65db420c643adfe2e457c)) ```sh lms get https://huggingface.co/Qwen/Qwen3-VL-32B-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 (~34.8 GB) `lms get https://huggingface.co/Qwen/Qwen3-VL-32B-Instruct-GGUF@Q8_0 -y` · 4bit (~19.6 GB, Mac only) `lms get https://huggingface.co/lmstudio-community/Qwen3-VL-32B-Instruct-MLX-4bit -y` · 8bit (~36 GB, Mac only) `lms get https://huggingface.co/lmstudio-community/Qwen3-VL-32B-Instruct-MLX-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 Qwen3-VL 32B Instruct (Q4_K_M, needs ~21 GB, from [ollama.com](https://ollama.com/library/qwen3-vl/tags)) ```sh ollama run qwen3-vl:32b-instruct ``` Other versions: Q8_0 (~36 GB) `ollama run qwen3-vl:32b-instruct-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 Qwen3-VL 32B Instruct (4bit, needs ~19.6 GB, Mac only, from [huggingface.co/lmstudio-community](https://huggingface.co/lmstudio-community/Qwen3-VL-32B-Instruct-MLX-4bit/tree/e95761a6100d3601c950823b233e0ce92576f94b)) ```sh mlx_lm.chat --model lmstudio-community/Qwen3-VL-32B-Instruct-MLX-4bit ``` 3. Or start an OpenAI-compatible server at http://localhost:8080/v1 ```sh mlx_lm.server --model lmstudio-community/Qwen3-VL-32B-Instruct-MLX-4bit ``` Other versions: 8bit (~36 GB, Mac only) `mlx_lm.chat --model lmstudio-community/Qwen3-VL-32B-Instruct-MLX-8bit` Model repository and documentation: [huggingface.co/Qwen](https://huggingface.co/Qwen/Qwen3-VL-32B-Instruct) [interactive version](https://whatmodel.sh/model/qwen3-vl-32b-instruct) · [JSON](https://whatmodel.sh/api?format=json&all=1)