Gemini grows into 3D

Google just turned Gemini from a chatbot into something closer to a little science museum inside a chat window. On April 9, 2026, Google said the Gemini app can now generate interactive simulations, 3D models, and charts instead of only text answers and flat diagrams. (blog.google) A text answer tells you that the Earth tilts 23.5 degrees. A 3D model lets you drag the planet, change the sun angle, and watch seasons happen like a desk lamp shining on a globe. (blog.google) Google’s example was education, not gaming. The company said Gemini can turn questions about topics like astronomy, biology, and geometry into custom visualizations that you can manipulate directly inside the conversation. (blog.google) That fits a longer shift inside Google’s model lineup. In December 2025, Google said Gemini 3 Pro was built for visual and spatial reasoning, meaning the model is trained to understand where things are, how shapes relate, and what changes when you rotate or move them. (blog.google) Google has also been pushing that same idea to developers, not just consumers. Its Gemini 3 documentation says the model family is meant for multimodal tasks, autonomous coding, and agentic workflows, and Google AI Studio now pitches Gemini as a tool for building apps from natural-language prompts and visual inputs. (ai.google.dev, developers.googleblog.com) The important change is not “Gemini can draw.” Google already had image and video tools like Nano Banana and Veo on its developer platform; this update puts interactive understanding into the main Gemini app, where a user can ask a question and get a model they can poke at, not just a picture they can stare at. (ai.google.dev, blog.google) That makes Gemini more useful for jobs where words are the wrong interface. If you are explaining a molecule, a supply chain, a bridge truss, or a probability curve, a movable object can carry the answer faster than five paragraphs of text. (blog.google, blog.google) The timing also lines up with a pricing fight over serious users. On April 9, 2026, OpenAI said it was adding a new $100-per-month Pro tier with 5 times more Codex usage than Plus, while keeping its existing $200 Pro plan as the top usage option. (community.openai.com, help.openai.com) Those two announcements point at the same customer from different angles. OpenAI is trying to capture developers who need longer coding sessions and higher limits, while Google is trying to make Gemini useful in classes, design work, and commercial software where understanding space, motion, and diagrams matters as much as writing code. (help.openai.com, ai.google.dev, blog.google) If this works, the assistant stops feeling like a search box with better manners. It starts acting more like a tool that can build a tiny working model on demand, which is a much stronger pitch to schools, developers, and businesses than “ask me anything.” (blog.google, blog.google)