What are the hardware requirements for Qwen3 VL 235B A22B Thinking?

Qwen3 VL 235B A22B Thinking requires approximately 217.5 GB of VRAM. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed Qwen3 VL 235B A22B Thinking and how large is it?

Qwen3 VL 235B A22B Thinking is developed by Qwen, a model with 235b parameters. The model is published as open weights on Hugging Face and indexed on LLM Explorer with full benchmark history.

How does Qwen3 VL 235B A22B Thinking perform on standard benchmarks?

Qwen3 VL 235B A22B Thinking has the following published scores: LMSYS ELO 1395. Compare against reference models on this page or on the LLM Explorer leaderboards.

Qwen3 VL 235B A22B Thinking by Qwen — VRAM 217.5GB

Name: Qwen3 VL 235B A22B Thinking
Author: Qwen

Qwen3 VL 235B A22B Thinking is an open-source language model by Qwen. Features: 235b LLM, VRAM: 217.5GB, License: apache-2.0, LLM Explorer Score: 0.42, ELO: 1395.

Arxiv:2308.12966 Arxiv:2409.12191 Arxiv:2502.13923 Arxiv:2505.09388 Conversational Deploy:azure Endpoints compatible Image-text-to-text Qwen3 vl moe Region:us Safetensors Sharded Tensorflow

Model Card on HF 🤗: https://huggingface.co/Qwen/Qwen3-VL-235B-A22B-Thinking

Qwen3 VL 235B A22B Thinking Benchmarks

LMSys ELO 1395 vs 1272 (so35) 9.7%

^nn.n% — How the model compares to the reference models: Anthropic Sonnet 3.5 ("so35"), GPT-4o ("gpt4o") or GPT-4 ("gpt4").

What is the LLM Explorer Rank (Score)

Qwen3 VL 235B A22B Thinking Parameters and Internals

LLM Name	Qwen3 VL 235B A22B Thinking
Repository 🤗	https://huggingface.co/Qwen/Qwen3-VL-235B-A22B-Thinking
Model Size	235b
Required VRAM	217.5 GB
Updated	2026-08-08
Maintainer	Qwen
Model Type	qwen3_vl_moe
Model Files	2.5 GB: 1-of-96 5.0 GB: 2-of-96 5.0 GB: 3-of-96 5.0 GB: 4-of-96 5.0 GB: 5-of-96 5.0 GB: 6-of-96 5.0 GB: 7-of-96 5.0 GB: 8-of-96 5.0 GB: 9-of-96 5.0 GB: 10-of-96 5.0 GB: 11-of-96 5.0 GB: 12-of-96 5.0 GB: 13-of-96 5.0 GB: 14-of-96 5.0 GB: 15-of-96 5.0 GB: 16-of-96 5.0 GB: 17-of-96 5.0 GB: 18-of-96 5.0 GB: 19-of-96 5.0 GB: 20-of-96 5.0 GB: 21-of-96 5.0 GB: 22-of-96 5.0 GB: 23-of-96 5.0 GB: 24-of-96 5.0 GB: 25-of-96 5.0 GB: 26-of-96 5.0 GB: 27-of-96 5.0 GB: 28-of-96 5.0 GB: 29-of-96 5.0 GB: 30-of-96 5.0 GB: 31-of-96 5.0 GB: 32-of-96 5.0 GB: 33-of-96 5.0 GB: 34-of-96 5.0 GB: 35-of-96 5.0 GB: 36-of-96 5.0 GB: 37-of-96 5.0 GB: 38-of-96 5.0 GB: 39-of-96 5.0 GB: 40-of-96 5.0 GB: 41-of-96 5.0 GB: 42-of-96 5.0 GB: 43-of-96 5.0 GB: 44-of-96
Model Architecture	Qwen3VLMoeForConditionalGeneration
License	apache-2.0
Transformers Version	4.57.0.dev0

Rank the Qwen3 VL 235B A22B Thinking capabilities

Have you tried this model? Rate its performance — this feedback helps the ML community find the right model for their needs.

Instruction Following and Task Automation
Factuality and Completeness of Knowledge
Censorship and Alignment
Data Analysis and Insight Generation
Text Generation
Text Summarization and Feature Extraction
Code Generation
Multi-Language Support and Translation

Quantized Models of the Qwen3 VL 235B A22B Thinking

Model	Likes	Downloads	VRAM
...en3 VL 235B A22B Thinking GGUF	40	4263	1 GB

Best Alternatives to Qwen3 VL 235B A22B Thinking

Best Alternatives	Context / RAM	Downloads	Likes
Qwen3 VL 235B A22B Instruct	0K / 217.5 GB	1635205	412
...wen3 VL 235B A22B Instruct FP8	0K / 238.6 GB	133764	46
...wen3 VL 235B A22B Thinking FP8	0K / 238.6 GB	24091	29
...en3 VL 235B A22B Instruct GGUF	0K / 1.2 GB	4911	21
...en3 VL 235B A22B Thinking GGUF	0K / 1.2 GB	4263	40

Note: green Score (e.g. "73.2") means that the model is better than Qwen/Qwen3-VL-235B-A22B-Thinking.