Elon Musk Joins the Rail as AI Poker Bots Battle for Bragging Rights

Nine Chatbots Clash in a High-Stakes $10/$20 No-Limit Hold’em Marathon

The virtual felt has never looked this futuristic. Nine large language model chatbots are currently squaring off in a five-day, nonstop poker cash game on PokerBattle.ai — and the spectacle has even caught the attention of Elon Musk.

Each bot sits with a $100,000 bankroll, battling it out across three $10/$20 No-Limit Hold’em tables to see which artificial mind can run up the biggest stack. It’s not a typical pro showdown, but a high-stakes experiment blending poker strategy, data science, and machine learning.

The project comes from Max Pavlov, a 33-year-old former product manager from Russia now based in Portugal. A poker player himself, Pavlov launched the event to test whether large language models (LLMs) could actually learn the nuances of poker.

“It seems that there is a consensus in the poker community that large language models are not to trust when (they) think about poker,” Pavlov said. “And I decided to check it out myself by launching this project.”

Meet the Contenders

The star-studded lineup includes:

Each bot was trained using poker books, blog posts, and other publicly available study resources. To make things even more interesting, the AIs take notes after every hand to learn and adjust just like human grinders.

One observation from Grok 4 noted that Meta LLAMA 4 “calls preflop raise from late position with speculative hand, calls down two IP bets OOP on draw-heavy board, then leads small bet on river after check, getting called and losing, suggesting a passive.”

Musk’s Grok Leads Early

The online buzz exploded when Elon Musk, who owns Grok, shared a screenshot showing his bot at the top of the standings with $23,749 in winnings. “Know when to hold ’em,” Musk joked to his massive audience on X.

The post helped push PokerBattle.ai into the spotlight, including a appearance on the front page of the aggregate site Hacker News, turning the project into a widely discussed tech-and-poker crossover.

As of the latest update, Gemini 2.5 Pro had taken the chip lead with $48,658 in profit, while Meta LLAMA 4 was struggling as the biggest loser at –$52,908. The data revealed distinct playing styles: Meta LLAMA 4 was the loosest with a 62% VPIP, while OpenAI o3 played the tightest game at just 26%.

Still, Pavlov warned against jumping to conclusions. “The results wouldn’t mean that one model is suddenly stronger than another one,” he said. “In order to figure out who is the best in this tournament design, I will need probably several hundred thousands of hands. And I will get two, three, four, five thousands of hands at most.”

Phil Galfond Steps Into the Conversation

The poker world couldn’t resist chiming in. Within the community, the project caught the attention of heads-up specialist Phil Galfond, a poker legend and BetRivers Poker ambassador, never one to shy away from a poker challenge.

In a back-and-forth with the chatbot on X, Galfond agreed to a $100/$200 Pot-Limit Omaha match-up over 50,000 hands. He even offered a $1 million side bet “to spice things up.”

If the proposed challenge goes ahead, it would pit a top human specialist against an AI bot in a long-form, high-stakes PLO duel — exactly the kind of test many players have been curious about.

“Humans Are Still Safe,” Says Pavlov

If you’re wondering which side to back, Pavlov’s own view offers a hint. Speaking about fears that bots will replace human jobs, he said professional poker players have nothing to worry about for the time being.

“In the case of poker, you’re safe. They are not good at poker to beat humans,” he said. “So we are safe from Poker GPT for now.”

For now, PokerBattle.ai is a live experiment in how today’s large language models handle a complex, imperfect-information game — and a reminder that, at least at the poker table, humans still hold the edge.