groundingpi / GroundingPI

English | 简体中文 
 A Grounding Foundation Model towards Physical Intelligence with Visual Primitives 
 Demo Video · Quick Start · Documentation · Citation 
## 📰 News
- **2026-10-14:** GroundingPI model weights are now available on ModelScope.
- **2026-10-14:** GroundingPI is featured in a report by 具身智能之心.
- **2026-10-14:** GroundingPI ranked **#1 Paper of the Day** on Hugging Face Daily Papers.
- **2026-10-04:** Released the project webpage.
- **2026-10-03:** Released the source code, deployment and batch-annotation guides, and full-suite evaluation workflows.
- **2026-10-01:** We released the GroundingPI model weights on Hugging Face.
- **2026-09-30:** The GroundingPI paper is available on arXiv.
## 🧭 Contents
Highlights · Demo · Models · Installation · Deployment Options · Quick Start · vLLM Deployment · Batch Annotation · Tasks and Output Format · Method and Inference Infrastructure · Evaluation · Training · Physical Intelligence · Results · Documentation · License · Citation · Acknowledgement
## ✨ Highlights
- **A strong grounding foundation model.** We introduce GroundingPI, a 4B model built on visual primitives, with a staged training recipe and state-of-the-art grounding performance.
- **Transfer toward physical intelligence.** Autonomous driving and robotic manipulation evaluations demonstrate the value of this perceptual foundation, including strong ID and OOD performance and improved action-data efficiency.
- **Insights into perceptual pretraining and future embodied paradigms.** We analyze how pretraining scale and data composition shape grounding and transfer, and discuss implications for System-1 foundation-model design and its complementary role in future embodied systems.
## 🎬 Demo
▶ Watch the demo
## 🤗 Models
| Checkpoint | Generation | Download | Evaluation mode |
|:---|:---|:---|:---|
| GroundingPI | Autoregressive visual grounding | Hugging Face | **GAM** |
The grounding checkpoint contains the vision-language model, tokenizer, processor, and custom model code. Downstream action-policy integrations are described in Physical Intelligence.
## 🛠️ Installation
The serving workflows below target **Linux x86_64 and Python 3.12**. Start from a compatible accelerator runtime with **Torch and vLLM 0.18.x already installed**. The serving setup inherits that runtime and installs the bundled **Transformers 5.7.0** fork.
 [code] 
Already have a source checkout? Start with `cd GroundingPI`. Run subsequent commands from the repository root.
`requirements.txt` installs the lightweight HTTP client, visualization tools, and setup dependencies. Serving, training, and evaluation each use their own environment; `pip install -r requirements.txt` alone does not install the model runtime. The client does not load weights and requires no Torch installation.
**Tested accelerators:** NVIDIA **B300, B200, H200, H800**, and **PPU**. Use the matching runtime for each accelerator. See Environment Setup for installation details.
## 🧩 Deployment Options
| Model