In Pictures: Xpeng unveils open-source GroundingPI perception model, claims edge over GPT-6 Astra on 34 benchmarks

Xpeng's GroundingPI model is designed to enhance perception for autonomous driving and robotics.

Photo by Kindel Media on Pexels

The open-source model aims to reduce the cost of collecting robot demonstration data.

Photo by Laura Musikanski on Pexels

GroundingPI scored 73.68 across 34 public perception benchmarks, surpassing GPT-6 Astra.

Photo by Alex Knight on Pexels

Developed with Peking University and University of Hong Kong researchers, GroundingPI targets embodied perception.

Photo by Pavel Danilyuk on Pexels

The model's open-source release could accelerate innovation in robotics and autonomous systems.

Photo by Pavel Danilyuk on Pexels

Xpeng claims GroundingPI outperforms OpenAI's GPT-6 Astra on 34 benchmarks.

Photo by Kindel Media on Pexels

GroundingPI is a roughly 4-billion-parameter foundation model for embodied perception.

Photo by Kindel Media on Pexels

The model's benchmark score of 73.68 tops GPT-6 Astra's 71.54.

Photo by Pavel Danilyuk on Pexels

Xpeng's open-source approach aims to cut costs in robot demonstration data collection.

Photo by Kindel Media on Pexels