GitLab Transcend - Intelligent orchestration
Mini toolsGamma World (γ-World)
By NVIDIA
Gamma-World is a generative multi-agent world model from NVIDIA, Tsinghua University, the University of Toronto, and the Vector Institute. It generates coherent future video frames from multiple agents’ action streams while preserving shared-world consistency, targeting multiplayer virtual environments and real-world multi-robot coordination. Its main technical components are Simplex Rotary Agent Encoding, which gives agents distinct but permutation-equivalent identities, and Sparse Hub Attention, which lets agents communicate efficiently while reducing cross-agent attention cost from quadratic to linear. The project reports real-time 24 FPS action-responsive rollouts and zero-shot generalization from two to four players without extra training.
New
Multimodal
Gen 3
Overview
Gamma-World, or γ-World, is NVIDIA’s generative multi-agent world model for real-time interactive video rollouts with independently controllable agents in a shared evolving environment.
About NVIDIA
Industry:
Computer Hardware Manufacturing
Company Size:
42000
Location:
Santa Clara, California, US
Website:
nvidia.com
Tools using Gamma World (γ-World)
No tools found for this model yet.
Last updated: June 2, 2026
