G²VLM
- AI model
- Perception
G²VLM is a vision-language model that predicts 3D geometry from images and answers spatial reasoning questions, combining a 3D reconstruction expert with a multimodal understanding expert.
Made by Intern Robotics, Shanghai AI Laboratory.
Source on GitHub: github.com/InternRobotics/G2VLM (opens in a new tab). Facts read at commit a45f2f4 (opens in a new tab).
Licenses
- Code
- Apache-2.0 (opens in a new tab)
Read by Robotheca from LICENSE (opens in a new tab) (Apache-2.0) at commit a45f2f4 (opens in a new tab), on 16 Oct 2026.
- Weights
- Apache-2.0 (opens in a new tab)
Read on huggingface.co/InternRobotics/G2VLM-2B-MoT (opens in a new tab).
- Commercial use
- Yes
What it does
- Performs 3D reconstruction from a folder of images and exports a point cloud as a .ply file.
- Answers spatial reasoning questions about images, using interleaved reasoning over predicted geometry.
- Uses a unified transformer where geometry and semantic tokens share multimodal self-attention.
- Provides command-line inference scripts for both 3D reconstruction and spatial question answering.
- Includes pre-training and joint-training scripts, with configurable datasets for fine-tuning.
- Pretrained checkpoint G2VLM-2B-MoT is downloadable from Hugging Face; a spatial reasoning checkpoint is announced as forthcoming.
Written by Robotheca from the project's README (opens in a new tab) at commit a45f2f4.
Getting started
- Difficulty
- Advanced
- How to start
- The project's README (opens in a new tab)
Maintenance
- Status
- Maintained: a commit in the last 12 months
- Newest commit
- 19 Apr 2026 on main
- Latest release
- None published on GitHub
Checked on GitHub on 16 Oct 2026.
Links
Robots that can run it
No robot on Robotheca is linked to it yet. Links come from a robot's robot.yaml, from its dependency files read as text, or from Robotheca's editors.
Robotheca lists AI models as a reference: nothing is hosted or run here, and every fact links to where it was read.