LLM EXPLORER 57,252 MODELS INDEXED

GELab Zero 4B Preview by stepfun-ai

By stepfun-ai · 235 downloads
HF Repo ↗ Copy URL

GELab Zero 4B Preview is an open-source language model by stepfun-ai. Features: 4b LLM, VRAM: 8.9GB, License: apache-2.0, Instruction-Based, LLM Explorer Score: 0.24.

  Arxiv:2512.02423   Arxiv:2512.15431   Android Base model:finetune:qwen/qwen3... Base model:qwen/qwen3-vl-4b-in...   Computer-use agent   Conversational   Deploy:azure   En   Endpoints compatible   Gelab-zero   Gui-agent   Image-text-to-text   Instruct   Multimodal   Phone-use agent   Pua   Qwen3 vl   Region:us   Safetensors   Sharded   Tensorflow   Zh
Model Card on HF 🤗: https://huggingface.co/stepfun-ai/GELab-Zero-4B-preview 

GELab Zero 4B Preview Parameters and Internals

LLM NameGELab Zero 4B Preview
Repository 🤗https://huggingface.co/stepfun-ai/GELab-Zero-4B-preview 
Base Model(s)  Qwen3 VL 4B Instruct   Qwen/Qwen3-VL-4B-Instruct
Model Size4b
Required VRAM8.9 GB
Updated2026-08-08
Maintainerstepfun-ai
Model Typeqwen3_vl
Instruction-BasedYes
Model Files  5.0 GB: 1-of-2   3.9 GB: 2-of-2
Supported Languagesen zh
Model ArchitectureQwen3VLForConditionalGeneration
Licenseapache-2.0
Model Max Length262144
Transformers Version4.57.1
Tokenizer ClassQwen2Tokenizer
Padding Token<|endoftext|>
Vocabulary Size151936
Errorsreplace

Best Alternatives to GELab Zero 4B Preview

Best Alternatives
Context / RAM
Downloads
Likes
Qwen3 VL 4B Instruct0K / 8.9 GB3611570437
Qwen3 VL 4B Instruct FP80K / 6.1 GB30110365
OMEGA 4B SpatialThink 08040K / 8.9 GB161
ZwZ 4B0K / 9.7 GB15532
Spatial SSRL Qwen3VL 4B0K / 9.7 GB10114
... VL 4B Instruct Uncensored FP80K / 5.7 GB20637
...L 4B Instruct Unsloth Bnb 4bit0K / 4.3 GB25381712
...en3 VL 4B Instruct Abliterated0K / 8.9 GB2383994
...wen3 VL 4B Instruct Uncensored0K / 17.8 GB1794
Qwen3 VL 4B Instruct 4bit0K / 3.1 GB174167
Note: green Score (e.g. "73.2") means that the model is better than stepfun-ai/GELab-Zero-4B-preview.