Google’s Gemini Models Lead AI Rankings in Social and Strategy Games

Google’s Gemini models are currently outperforming rivals in social and strategic games. Google DeepMind, together with Kaggle, has expanded its Game Arena benchmark with two new titles: Werewolf and Poker. The platform is designed to evaluate AI models through competitive games that test different cognitive skills.

• AI Safety
• Gemini
• AI Benchmarks
• AI models
• DeepMind

Each game targets a distinct capability. Chess measures logical reasoning, Werewolf evaluates social intelligence such as communication, deception detection, and theory of mind, while Poker tests decision-making under uncertainty, risk management, and incomplete information.

According to the latest results, Gemini 3 Pro and Gemini 3 Flash currently top all leaderboards across the Game Arena benchmarks. The Werewolf benchmark also plays a role in AI safety research, as it allows researchers to assess whether models can detect manipulation and deceptive behavior without exposing them to real-world risks.

Google DeepMind CEO Demis Hassabis said the results highlight the need for more demanding and realistic evaluations of next-generation AI systems, arguing that the industry requires tougher benchmarks to properly assess emerging model capabilities.

Chris Borden

AI Analyst & Technology Researcher

AI researcher and industry analyst covering decentralized infrastructure, AI systems, and emerging technology markets. Focused on data-driven analysis, long-term trends, and real-world adoption of artificial intelligence.