What are the hardware requirements for AREX Turbo?

AREX Turbo requires approximately 9.1 GB of VRAM. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed AREX Turbo and how large is it?

AREX Turbo is developed by BAAI, a model with 4b 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 AREX Turbo?

AREX Turbo is hosted on Hugging Face and linked from this page. LLM Explorer also lists quantized variants and similar alternatives if available.

AREX Turbo by BAAI — VRAM 9.1GB

Name: AREX Turbo
Author: BAAI

AREX Turbo is an open-source language model by BAAI. Features: 4b LLM, VRAM: 9.1GB, License: apache-2.0, LLM Explorer Score: 0.34.

Arxiv:2607.21461 Agent Base model:finetune:qwen/qwen3... Base model:qwen/qwen3.5-4b Conversational Deep-research Dense En Endpoints compatible Image-text-to-text Long-context Qwen3.5 Qwen3 5 Reasoning Region:us Safetensors Sharded Tensorflow Tool-use

Model Card on HF 🤗: https://huggingface.co/BAAI/AREX-Turbo

AREX Turbo Parameters and Internals

LLM Name	AREX Turbo
Repository 🤗	https://huggingface.co/BAAI/AREX-Turbo
Base Model(s)	Qwen/Qwen3.5-4B Qwen/Qwen3.5-4B
Model Size	4b
Required VRAM	9.1 GB
Updated	2026-08-10
Maintainer	BAAI
Model Type	qwen3_5
Model Files	5.0 GB: 1-of-2 4.1 GB: 2-of-2
Supported Languages	en
Model Architecture	Qwen3_5ForConditionalGeneration
License	apache-2.0
Model Max Length	262144
Transformers Version	5.2.0
Tokenizer Class	Qwen2Tokenizer
Padding Token	<\|endoftext\|>
Errors	replace

Rank the AREX Turbo 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 AREX Turbo

Model	Likes	Downloads	VRAM
AREX Turbo 4bit	1	74	3 GB
AREX Turbo 8bit	1	63	5 GB
AREX Turbo 6bit	1	41	4 GB

Best Alternatives to AREX Turbo

Best Alternatives	Context / RAM	Downloads	Likes
Qwen3.5 4B	0K / 9.3 GB	6575502	801
Agents A1 4B	0K / 9.1 GB	177578	72
NuExtract3	0K / 0.2 GB	441570	324
Fara1.5 4B	0K / 9.1 GB	4934	39
Qwen3.5 4B Base	0K / 9.3 GB	227814	79
DeepSeek V4 Pro Qwen3.5 4B	0K / 9.3 GB	3476	9
JOSIE 2 4B Preview	0K / 9.3 GB	733	20
Mephisto 4B 0725	0K / 8.6 GB	3001	5
Qwen3.5 4B	0K / 9.3 GB	139961	26
JOSIE 2 4B OSS	0K / 9.3 GB	325	7

Note: green Score (e.g. "73.2") means that the model is better than BAAI/AREX-Turbo.