LLM EXPLORER 58,178 MODELS INDEXED

Qwen3 VL 30B A3B Thinking 4bit by mlx-community

By mlx-community · 70 downloads
HF Repo ↗ Copy URL

Qwen3 VL 30B A3B Thinking 4bit is an open-source language model by mlx-community. Features: 30b LLM, VRAM: 18.3GB, License: apache-2.0, Quantized, LLM Explorer Score: 0.21.

  4-bit   4bit   Conversational   Image-text-to-text   Mlx   Quantized   Qwen3 vl moe   Region:us   Safetensors   Sharded   Tensorflow
Model Card on HF 🤗: https://huggingface.co/mlx-community/Qwen3-VL-30B-A3B-Thinking-4bit 

Qwen3 VL 30B A3B Thinking 4bit Parameters and Internals

LLM NameQwen3 VL 30B A3B Thinking 4bit
Repository 🤗https://huggingface.co/mlx-community/Qwen3-VL-30B-A3B-Thinking-4bit 
Model Size30b
Required VRAM18.3 GB
Updated2026-08-12
Maintainermlx-community
Model Typeqwen3_vl_moe
Model Files  5.3 GB: 1-of-4   5.4 GB: 2-of-4   5.3 GB: 3-of-4   2.3 GB: 4-of-4
Quantization Type4bit
Model ArchitectureQwen3VLMoeForConditionalGeneration
Licenseapache-2.0
Model Max Length262144
Transformers Version4.57.0
Padding Token<|endoftext|>
Errorsreplace

Best Alternatives to Qwen3 VL 30B A3B Thinking 4bit

Best Alternatives
Context / RAM
Downloads
Likes
Qwen3 VL 30B A3B Instruct 4bit0K / 18.3 GB9258
Bu 30B A3b Preview 4bit0K / 18.3 GB524
Qwen3 VL 30B A3B Instruct 8bit0K / 33.5 GB2313
Qwen3 VL 30B A3B Instruct 6bit0K / 25.8 GB722
Qwen3 VL 30B A3B Thinking 3bit0K / 14.5 GB401
Qwen3 VL 30B A3B Instruct0K / 62.3 GB488106591
Qwen3 VL 30B A3B Instruct FP80K / 32.2 GB734631115
Qwen3 VL 30B A3B Instruct AWQ0K / 17.9 GB118470445
Qwen3 VL 30B A3B Thinking0K / 62.3 GB67001200
Qwen3 VL 30B A3B Thinking FP80K / 32.2 GB4300458
Note: green Score (e.g. "73.2") means that the model is better than mlx-community/Qwen3-VL-30B-A3B-Thinking-4bit.