Google’s newest AI breakthrough, Gemini 3.1 Pro, is the new number one on APEX-Agents leaderboard. This truly mindboggling feat underscores just how powerful and extraordinary it truly is that AI can perform feats of actual knowledge work. Google shared numbers Thursday that illustrate just how big a leap in performance the new model represents. These improvements were very noticeable even in unbiased benchmarks, such as one notoriously titled Humanity’s Last Exam.
This proficiency is a hallmark of the Gemini 3.1 Pro, a Large Language Model (LLM) built from the ground up to do agentic work and multi-step reasoning. At the moment, it’s available as a limited preview, but a broader launch is projected to come in the near future. AI capabilities are advancing at a breakneck pace. Today, we see tech companies brutally competing to create even more advanced LLMs.
At an event held in Boston, MA, on June 9, 2026, Google representatives discussed the advancements of Gemini 3.1 Pro. Brendan Foody emphasized the model’s achievements, stating, “Gemini 3.1 Pro is now at the top of the APEX-Agents leaderboard.” He called attention to the rapid pace of agents advancing their capabilities in truly knowledge work. This highlights the extraordinary pace of change occurring within the AI industry.
Independent benchmarks have shown that Gemini 3.1 Pro is far more powerful than its predecessor. As the generative AI model wars heat up, this model’s release places it as one of the strongest contenders among its fellow competitors.
Today’s release of Gemini 3.1 Pro should be an exciting time for Google. It’s an indicator of the broader trends in AI development, as companies race to be the first to push the boundaries of machine learning and natural language processing. The buzz around its general release underscores the rapidly increasing excitement about these LLMs and their powerful uses across many different domains.
