Gemini 3.1 Pro (Preview) (Google Vertex AI)

Gemini 3.1 Pro Preview is positioned by Google as a refinement of the Gemini 3 Pro family, with the preview release emphasizing stronger thinking, better token efficiency, and a more grounded, factually consistent response style. It is tuned for software engineering behavior and usability, and is described as optimized for agentic workflows that demand precise tool usage and reliable multi-step execution across real-world domains. The model stays in the proprietary Gemini line rather than moving to an open-weight release, and it is delivered through Google Vertex AI for production access.

In practice, the preview offers a roughly one-million-token that quick-info value window with a 65,536-token output ceiling, accepting text, image, audio, video, and PDF inputs while returning text, which makes it well suited to long-document analysis, code and repository reasoning, and multimodal assistants. Catalog sources frame it as a follow-up that consolidates the Gemini 3 Pro series' strengths, targeting developers who need dependable function calling, structured output, and temperature control for agent pipelines. For teams already invested in Vertex AI, it serves as the natural successor for coding copilots, research synthesis, and multi-step automation that benefits from extended context.