Google DeepMind launches Gemini 3 AI with agentic features, delays Gemini 3.5 Pro to July 17

Google DeepMind released Gemini 3 Pro on 6 July 2026, offering sharper logical reasoning, multi‑step task automation and a redesigned interface. The model adds agentic capabilities that let it coordinate workflows across Google applications, and introduces a “Thinking” mode for advanced logical processing. Higher‑tier subscribers (AI Plus, Pro, Ultra) receive raised usage limits, and US college students get a free year of Google AI Pro. An experimental Gemini Agent, available to Ultra users in the US, executes multi‑stage tasks by directly interacting with other Google services.

A macOS beta, Gemini Spark, debuted a day earlier, providing a 24‑hour personal AI assistant that can sort files, update documents and create spreadsheets on command, with integrations for Canva, Dropbox, Instacart, OpenTable and Zillow Rentals. New specialist models such as Nano Banana 2 Lite and the Gemini Omni Flash preview were also announced.

Separately, Google DeepMind postponed the Gemini 3.5 Pro launch from June to 17 July 2026 to extend pre‑training. The delay targets improvements in mathematical reasoning, SVG scene generation and image generation, aiming to keep the model competitive with OpenAI’s upcoming GPT‑5.6 and Anthropic’s Fable 5.