LLM EXPLORER 62,479 MODELS INDEXED

AREX Turbo by BAAI

By BAAI · 358 downloads
HF Repo ↗ Copy URL

AREX Turbo is an open-source language model by BAAI. Features: 4b LLM, VRAM: 9.1GB, License: apache-2.0, LLM Explorer Score: 0.29.

  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 NameAREX Turbo
Repository 🤗https://huggingface.co/BAAI/AREX-Turbo 
Base Model(s)  Qwen/Qwen3.5-4B   Qwen/Qwen3.5-4B
Model Size4b
Required VRAM9.1 GB
Updated2026-09-29
MaintainerBAAI
Model Typeqwen3_5
Model Files  5.0 GB: 1-of-2   4.1 GB: 2-of-2
Supported Languagesen
Model ArchitectureQwen3_5ForConditionalGeneration
Licenseapache-2.0
Model Max Length262144
Transformers Version5.2.0
Tokenizer ClassQwen2Tokenizer
Padding Token<|endoftext|>
Errorsreplace

Quantized Models of the AREX Turbo

Model
Likes
Downloads
VRAM
AREX Turbo 4bit1743 GB
AREX Turbo 8bit1635 GB
AREX Turbo 6bit1414 GB

Best Alternatives to AREX Turbo

Best Alternatives
Context / RAM
Downloads
Likes
Qwen3.5 4B0K / 9.3 GB6575502801
Intern Decision 4B0K / 0.1 GB33848
NuExtract30K / 0.2 GB441570324
Agents A1 4B0K / 9.1 GB17757872
Tev1 4B Experimental0K / 9.3 GB153916
Qwen3.5 4B Base0K / 9.3 GB22781479
APUS OpenJev V1 4B0K / 9.1 GB6654
CogEvol 4B0K / 9.1 GB87927
Fara1.5 4B0K / 9.1 GB493439
Qwen3.8 4B Distill0K / 9.3 GB187826
Note: green Score (e.g. "73.2") means that the model is better than BAAI/AREX-Turbo.