- OpenAI's o3 model won a five-day poker tournament with nine AI chatbots
- The o3 model won by playing the most consistent game
- Most of the best language models handled poker well, but struggled with bluffing, position, and basic math.
In a digital showdown unlike anything ever seen on the felt, nine of the world's most powerful language majors spent five days locked in a high-stakes poker game.
OpenAI's O3, Anthropic's Claude Sonnet 4.5, X.ai's Grok, Google's Gemini 2.5 Pro, Meta's Llama 4, DeepSeek R1, Moonshot AI's Kimi K2, Mistral AI's Magistral and Z.AI's GLM 4.6 played thousands of hands of no-limit Texas Hold'em at $10 and $20 tables with bankrolls of $100,000. each.
When OpenAI's o3 model walked away from a week-long poker game winning $36,691, there was no trophy, just bragging rights.
The experimental PokerBattle.ai was run entirely with AI and the same initial message was sent to each player. It was pure strategy, if strategy is what is called thousands of micro-decisions made by machines that don't really understand winning, losing or how humiliating it is to go over with the seven-two.
For a technological hack, it was unusually revealing. The highest-performing AIs didn't just bluff and gamble: they adapted, modeled their opponents, and learned in real time to navigate ambiguity. While they didn't play flawless poker, they came impressively close to imitating the decisions of experienced players.
OpenAI's o3 quickly demonstrated that he had the most stable hand, taking three of the five largest pots and approaching textbook pre-flop theory. Anthropic's Claude and X.com's Grok rounded out the top three with substantial earnings of $33,641 and $28,796, respectively.
Meanwhile, Llama lost his entire stack and burned out early. The rest of the field fell somewhere in the middle, with Google's Gemini making a modest profit and Moonshot's Kimi K2 losing chips to a finish of $86,030.
Gambling AI
Poker has long been one of the best analogs for testing general-purpose AI. Unlike chess or Go, which rely on perfect information, poker requires players to reason under conditions of uncertainty. It's a mirror of real-world decision-making in everything from business negotiations to military strategy and now, apparently, chatbot development.
A consistent finding from the tournament was that the robots were often too aggressive. Most favored high-action strategies, even in situations where retreating would have been smarter. They tried to win big pots rather than avoid losing them. And they were terrible at bluffing, not because they didn't try, but because their bluffing often came from misinterpreted hands, not clever deception.
Still, AI tools are getting smarter in ways that go far beyond surface intelligence. They are not just repeating what they have read; They are making probabilistic judgments under pressure and learning to read the room. It's also a reminder that even the most powerful models still have flaws. Misinterpreting situations, drawing shaky conclusions and forgetting your own “position” is not just a poker problem.
You may never sit in front of a language model in a real poker room, but chances are you will interact with one trying to make important decisions. This game was just a glimpse of what that could be like.
Follow TechRadar on Google News and add us as a preferred source to receive news, reviews and opinions from our experts in your feeds. Be sure to click the Follow button!
And of course you can also follow TechRadar on TikTok for news, reviews, unboxings in video form and receive regular updates from us on WhatsApp also.
The best business laptops for every budget
