Allen Institute of Artificial Intelligence

Tülu 3 405B, the latest in the Tülu 3 series, outperforms DeepSeek-V3, rivals GPT-4o & other open-weight post-trained models like Llama 3.1. Leveraging Reinforcement Learning from Verifiable Rewards (RVLR), it scales to 405B parameters, setting new benchmarks.

Objaverse is A massive dataset with 800k+ annotated 3D objects. Objaverse improves upon present day 3D repositories in terms of scale, number of categories, and in the visual diversity of instances within a category.