Google has unveiled Gemini Pro 3.1, the latest iteration of its large language model, marking a significant leap in performance and capabilities. The updated version is currently available as a preview and is expected to roll out more widely in the near future, according to the company. Gemini 3.1 Pro has already drawn attention for its standout results on a range of independent benchmarks. Notably, it outperforms its predecessor, Gemini 3, which was released in November and was already considered a top-tier AI model. The latest version has achieved record scores on evaluations such as Humanity’s Last Exam, a rigorous test designed to assess AI systems’ ability to handle complex, real-world challenges. Brendan Foody, CEO of AI startup Mercor, praised the model’s performance on his company’s benchmarking platform, APEX, which evaluates AI agents on their ability to perform professional-level tasks. “Gemini 3.1 Pro is now at the top of the APEX-Agents leaderboard,” Foody announced on social media. He highlighted that the results underscore how rapidly AI agents are advancing in their capacity to handle sophisticated knowledge work, including planning, research, and decision-making. The release of Gemini 3.1 Pro comes amid a growing competition in the AI space, as major tech firms race to deliver more capable, agentic models. Google’s move follows recent launches from competitors like OpenAI and Anthropic, all of which are pushing the boundaries of multi-step reasoning, tool use, and autonomous task execution. With this update, Google is reinforcing its position in the race to build the most powerful and practical AI systems for real-world applications.