Runwayml has released Praxis-1, a 27-billion parameter mixture-of-experts diffusion transformer model designed for predicting video and actions. This model has achieved top performance on NVIDIA's RoboLab-120 leaderboard, surpassing other open-weight video-language action models. Praxis-1 is available under an Apache 2.0 license on GitHub and Hugging Face, supporting commercial use and running on a range of NVIDIA hardware. AI
IMPACT Sets new SOTA on RoboLab-120 benchmark for video-language action models.
RANK_REASON Frontier-lab model release with system card [lever_c_demoted from frontier_release: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →