Google Launches Gemini 3.1 Pro on Vertex AI
Google has released Gemini 3.1 Pro in preview on its Vertex AI platform, featuring a 1 million token context window and multimodal support for audio, image, and video. The new model is designed for complex, agentic workflows requiring enhanced reasoning. The release coincides with a growing focus on deploying large language models locally on mobile devices, with new guides emerging on how to run LLMs on Android-based endpoints.
- The 1 million token context window allows Gemini 3.1 Pro to process and analyze vast amounts of information in a single prompt, equivalent to approximately 1 hour of video, 11 hours of audio, or codebases with over 30,000 lines. This capability is a significant increase from the 32,000 tokens supported by Gemini 1.0. - Gemini 3.1 Pro demonstrates improved reasoning capabilities, scoring 77.1% on the ARC-AGI-2 benchmark, which evaluates the model's ability to solve new logic patterns—more than double the performance of the previous Gemini 3 Pro. This is particularly relevant for complex, multi-step agentic workflows in areas like supply chain optimization and warehouse automation. - The model's multimodal capabilities natively handle text, images, audio, video, and code, allowing for sophisticated analysis of diverse data types. For logistics, this could mean simultaneously analyzing warehouse camera feeds, inventory databases, and shipping manifests to identify operational inefficiencies. - Google is positioning its Vertex AI platform, which hosts Gemini 3.1 Pro, as a comprehensive environment for building and deploying enterprise-grade AI agents. This includes tools like the Agent Development Kit (ADK) and Agent Engine to manage the entire lifecycle of these autonomous systems. - For on-device applications, Google is developing Gemini Nano, a more lightweight model designed to run directly on mobile endpoints like Android devices. This aligns with edge computing strategies, enabling AI-powered features such as on-device scam detection in phone calls. - Gemini 3.1 Pro introduces enhanced function calling and the ability to generate website-ready animated SVGs directly from text prompts. These code-based animations are scalable and have smaller file sizes than traditional video formats. - The release is part of Google's broader "Gemini Era," which involves integrating its AI models across its entire product suite, including Workspace, Google Search, and Android Studio, to enhance productivity and developer efficiency. - To manage costs associated with the large context window, Vertex AI offers context caching, allowing developers to reuse frequently accessed tokens across multiple prompts at a lower cost.
