Google's new Gemini Pro model has record benchmark scores—again | TechCrunch

TechCrunch

by Lucas Ropek

February 20, 2026

AI-Generated Deep Dive Summary

Google has unveiled Gemini 3.1 Pro, its latest large language model (LLM), which has achieved record-breaking performance on independent benchmarks. The updated version of Gemini Pro, available as a preview and set for broader release soon, demonstrates significant improvements over its predecessor, Gemini 3, which was already lauded for its advanced capabilities when released in November. Independent tests, including Humanity’s Last Exam, highlight the model’s superior performance, with CEO of AI startup Mercor Brendan Foody calling it the top performer on the APEX-Agents leaderboard. Foody emphasized the rapid progression of AI agents in handling complex professional tasks. Gemini 3.1 Pro’s advancements come amid intensifying competition in the AI sector, as major players like OpenAI and Anthropic also release new models focused on agentic work and multi-step reasoning. These developments reflect a race to create more powerful tools capable of tackling real-world challenges with greater efficiency and accuracy. The model’s improvements are particularly notable in its ability to perform knowledge work, a critical factor for businesses looking to integrate advanced AI into their operations. The release underscores the accelerating pace of innovation in AI technology. As tech companies continue to push the boundaries of what LLMs can do, models like Gemini 3.1 Pro represent significant strides toward more capable and practical applications. For readers interested in tech and AI advancements, this development highlights how quickly AI is evolving and its potential to transform industries by automating complex tasks and enhancing decision-making processes.

Verticals

techstartups

Originally published on TechCrunch on 2/20/2026