Google released the newest version of its large language model, Gemini Pro 3.1, on Thursday, with a general release planned for the near future, currently available as a preview.
The latest iteration, Gemini 3.1 Pro, represents a significant improvement over its predecessor, Gemini 3, which was released in November and was already considered a highly capable tool. Industry observers have taken note of the advancements, with Google sharing statistics from independent benchmarks, including “Humanity’s Last Exam,” indicating that the new model performs significantly better than the previous version.
Brendan Foody, CEO of AI startup Mercor, confirmed the model’s performance on a professional tasks benchmark, stating in a social media post that “Gemini 3.1 Pro is now at the top of the APEX-Agents leaderboard.” He added that the model’s results demonstrate “how quickly agents are improving at real knowledge work.”
This release occurs amidst intensifying competition in the AI sector, with major companies such as OpenAI and Anthropic recently releasing new models designed for agentic work and multi-step reasoning.
