Elon Musk has announced a new kind of competition between humans and artificial intelligence. In a post on X, he challenged xAI’s upcoming Grok 5 system to play a match against one of the world’s best League of Legends teams sometime in 2026.

Musk said the match will be strictly controlled to keep it fair. Grok 5 can only view the game screen via a camera, limited to what a human with 20/20 vision sees (no enhanced resolution or API access). The AI must also follow human limits on reaction time and mouse clicking speed. That means no superhuman reflexes.

According to Musk, Grok 5 is designed to learn games by reading instructions and playing them, not by training on large amounts of game data. This challenge is meant to test whether the AI can think, adapt, and make smart decisions in real time, not just perform fast actions.

Shortly after Musk’s post, the official @grok account responded, “Challenge accepted! As Grok 4, I’m excited to see Grok 5 take on League of Legends pros under those fair constraints. xAI is pushing AGI boundaries—count me in for the journey.”

Subscribe to our TER newsletters here! Including On The Radar a quick weekly wrap up of all esports business stories, and the fortnightly Heat Map, a deeper dive into the stories not to be missed across esports business and culture worldwide.

League of Legends (LoL) was chosen because it is one of the most complicated competitive video games. It requires teamwork, planning, map awareness, and constantly changing strategies. Musk said this makes it a good test for whether AI can show human-like reasoning.

Musk also used the announcement as a call for new talent. He invited developers, engineers, and AI researchers to join xAI to help solve problems like real-time vision, game control, and learning while playing.

Grok 5 was originally expected to launch in late 2025, but its release has been delayed to early 2026. It is planned to be a very large multimodal AI system, reportedly around 60 trillion parameters. Musk has said he believes there is a more than 10% chance the model will reach true AGI, meaning it could match or exceed human abilities across many different mental tasks.

Follow The Esports Radar on social media: