What are the hardware requirements for Kimi VL A3B Thinking 4bit?

Kimi VL A3B Thinking 4bit requires approximately 9.9 GB of VRAM. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed Kimi VL A3B Thinking 4bit and how large is it?

Kimi VL A3B Thinking 4bit is developed by mlx-community, a model with 0.5b parameters. The model is published as open weights on Hugging Face and indexed on LLM Explorer with full benchmark history.

Where can I download or evaluate Kimi VL A3B Thinking 4bit?

Kimi VL A3B Thinking 4bit is hosted on Hugging Face and linked from this page. LLM Explorer also lists quantized variants and similar alternatives if available.

Kimi VL A3B Thinking 4bit by mlx-community — VRAM 9.9GB

Name: Kimi VL A3B Thinking 4bit
Author: mlx-community

Kimi VL A3B Thinking 4bit is an open-source language model by mlx-community. Features: 0.5b LLM, VRAM: 9.9GB, License: other, Quantized, LLM Explorer Score: 0.19.

4bit Base model:merge:opengvlab/int... Base model:merge:qwen/qwen2.5-... Base model:opengvlab/internvit... Base model:qwen/qwen2.5-0.5b Conversational Custom code Dataset:opengvlab/mmpr-v1.2 Feature-extraction Image-text-to-text Internvl Kimi vl Mlx Multilingual Quantized Region:us Safetensors Sharded Tensorflow

Model Card on HF 🤗: https://huggingface.co/mlx-community/Kimi-VL-A3B-Thinking-4bit

Kimi VL A3B Thinking 4bit Parameters and Internals

LLM Name	Kimi VL A3B Thinking 4bit
Repository 🤗	https://huggingface.co/mlx-community/Kimi-VL-A3B-Thinking-4bit
Base Model(s)	InternViT 300M 448px V2.5 Qwen/Qwen2.5-0.5B OpenGVLab/InternViT-300M-448px-V2_5 Qwen/Qwen2.5-0.5B
Model Size	0.5b
Required VRAM	9.9 GB
Updated	2026-07-21
Maintainer	mlx-community
Model Type	kimi_vl
Model Files	5.4 GB: 1-of-2 4.5 GB: 2-of-2
Quantization Type	4bit
Model Architecture	KimiVLForConditionalGeneration
License	other
Model Max Length	1048576
Transformers Version	4.52.0.dev0
Tokenizer Class	TikTokenTokenizer
Padding Token	[PAD]
Vocabulary Size	163840
Torch Data Type	bfloat16

Rank the Kimi VL A3B Thinking 4bit 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