Google’s new Gemini Pro model has record benchmark scores — again
Google released Gemini 3.1 Pro, which achieved record benchmark scores, surpassing its predecessor, Gemini 3.Summary
Google recently released the newest version of its powerful large language model (LLM), Gemini 3.1 Pro, currently available as a preview. This new model represents a significant advancement over its predecessor, Gemini 3. Independent benchmark tests, including one named Humanity’s Last Exam, showed Gemini 3.1 Pro performing substantially better than previous versions. Furthermore, Brendan Foody, CEO of AI startup Mercor, confirmed that Gemini 3.1 Pro topped the APEX-Agents leaderboard, a system measuring performance on real professional tasks, indicating rapid improvement in AI agents for knowledge work. This release occurs amidst intensifying competition among tech giants like OpenAI and Anthropic in developing increasingly capable LLMs for agentic work and multi-step reasoning.
(Source：TechCrunch)