On Thursday, Google unveiled the latest iteration of its flagship large language model, Gemini 3.1 Pro. Currently available in preview, the company said a broader public rollout is coming soon.

 

Early reactions suggest Gemini 3.1 Pro could be among the most powerful LLMs to date. Observers have described it as a significant leap forward from Gemini 3, which was already widely regarded as a highly capable AI system when it launched in November.

 

Google also published results from independent benchmarks — including one known as Humanity’s Last Exam — indicating that the new model substantially outperforms its predecessor.

 

Industry leaders have taken note. Brendan Foody, CEO of AI startup Mercor, said Gemini 3.1 Pro now leads the APEX-Agents leaderboard, part of Mercor’s APEX benchmarking system, which evaluates how effectively AI models handle real-world professional tasks. In a social media post, Foody said the results highlight “how quickly agents are improving at real knowledge work.”

 

The launch arrives amid intensifying competition in the AI sector, as major players race to develop more advanced models capable of agentic behavior and complex, multi-step reasoning. Companies such as OpenAI and Anthropic have also rolled out new models in recent months.