Gemini 3.1 Pro Preview – Vertex AI – Google Cloud console

Google has released Gemini 3.1 Pro Preview on Vertex AI, positioning it as their most powerful agentic and coding model. It features a 1M token input context window, 64k output token limit, and supports audio, images, video, text, and PDF inputs. Key changes from Gemini 2.5 Pro include a new thinking level parameter (Low/Medium/High) replacing the legacy thinking_budget, stricter validation for function calling (removed thought signatures cause 400 errors), multimodal function response support, streaming function calling, and updated media resolution defaults for image, video, and PDF modalities.