Google Pushes Gemini 3.1 Pro to Enterprise

Gemini 3.1 Pro marks a significant upgrade in core reasoning, doubling its performance over Gemini 3 Pro on benchmarks like ARC-AGI-2, which tests for solving novel logic patterns. This enhancement targets complex, multi-step tasks where simple answers are insufficient, improving its ability to synthesize data and explain intricate topics. The model is rolling out across consumer and developer products, including the Gemini API, Vertex AI, and the Gemini app. The new model introduces a massive 1 million token context window, a significant leap from the 32.8K tokens of Gemini 1.0 Pro. This allows it to process vast amounts of information from various sources simultaneously, including text, images, audio (up to 8.4 hours), video (up to an hour without audio), and entire code repositories. The output token limit has also been expanded to 65,536, addressing a previous limitation where code generation would be truncated. For developers, Gemini 3.1 Pro is optimized for software engineering and "agentic" workflows—tasks requiring autonomous, multi-step execution. It shows improved instruction following, more reliable tool use, and better performance on coding benchmarks like SWE-Bench, achieving an 80.6% pass rate in resolving real-world issues from Python repositories. A separate endpoint is even provided for workflows mixing custom tools and bash scripts. Vertex AI's MLOps tools are designed to support the entire lifecycle of generative AI models like Gemini. The platform provides tools for every stage, from model discovery in the Model Garden and prompt tuning to managing model versions in the Model Registry and monitoring for drift in production. This allows teams to automate, standardize, and govern their ML projects, whether using Google's foundation models or open-source alternatives.