OpenAI is holding a last live event for OpenAI Five, the AI project that once put reinforcement learning on the esports map.
The event is scheduled for April 13 at 11:30am PT. The announcement said little beyond the time and the word "final" — no format, no guests, no agenda. OpenAI Five was built to play Dota 2, a complex multiplayer game that the lab used to benchmark what scaled reinforcement learning could accomplish. Its most notable moments came when the system faced professional players in public matches that drew real attention to what AI agents could do in adversarial, real-time environments.
Shutting the door on OpenAI Five makes sense in context. The lab's focus has shifted almost entirely to large language models and the commercial products built on them. Game-playing AI, once a proving ground for coordination and long-horizon reasoning research, isn't where the money or the attention is anymore.
Expect a retrospective, not a rematch. The game has moved on, and so has OpenAI.