OpenAI’s o3 model triumphed in a groundbreaking five-day poker tournament against nine AI chatbots, boasting a profit of $36,691. The competition featured top models including Anthropic’s Claude Sonnet 4.5 and Google’s Gemini 2.5 Pro, all engaging in no-limit Texas hold ’em with sizable bankrolls. While these AIs demonstrated remarkable strategic adaptations and real-time learning, they struggled with bluffing and made errors in basic math, often over-aggressively chasing big pots instead of playing conservatively. The tournament illustrated the complexities of decision-making under uncertainty, highlighting both the strengths and weaknesses of AI in poker and beyond. Despite their advanced algorithms, models like Llama experienced significant losses, while o3’s consistent gameplay earned it top honors. This event not only showcases AI capabilities but also serves as a reminder of the challenges they face in real-world applications. Follow TechRadar for updates on AI trends, reviews, and technology insights.
Source link
