---
license: apache-2.0
base_model: OpenGVLab/InternVL3-8B
pipeline_tag: image-text-to-text
library_name: transformers
tags:
- openspatial
- spatial-reasoning
- vision-language
---

# OpenSpatial-InternVL3-8B

OpenSpatial-InternVL3-8B is an OpenSpatial model fine-tuned for spatial understanding and
reasoning. This repository contains full, inference-ready model weights.

- Project: https://github.com/VINHYU/OpenSpatial
- Paper: https://arxiv.org/abs/2604.07296
- Training data: https://huggingface.co/datasets/jdopensource/JoyAI-Image-OpenSpatial
- Base model: https://huggingface.co/OpenGVLab/InternVL3-8B

## Loading

Use `transformers` the version recommended by InternVL3 for this checkpoint.

```python
from transformers import AutoModel, AutoTokenizer

model_id = "VINHYU/OpenSpatial-InternVL3-8B"
model = AutoModel.from_pretrained(
    model_id,
    torch_dtype="auto",
    low_cpu_mem_usage=True,
    trust_remote_code=True,
).eval().cuda()
tokenizer = AutoTokenizer.from_pretrained(
    model_id, trust_remote_code=True, use_fast=False
)
```

Please refer to the OpenSpatial repository and the base model repository for
full usage instructions, limitations, and license terms.