Google's Gemini 2.5 Pro

Google has widened what its Gemini Pro tier can do, adding interactive 3D models in the Gemini app and rolling out a heavier-duty Deep Research Max mode for longer research jobs. (blog.google 1) (blog.google 2) The 3D feature arrived on April 9, when Google said Gemini could generate “interactive simulations and models” inside chat for prompts like molecules, physics systems, and orbital mechanics. Google said the feature is rolling out globally in the Gemini app and requires selecting the Pro model in the prompt bar. (blog.google) Deep Research Max followed on April 21 as a public preview in the paid tiers of the Gemini Application Programming Interface, or API. Google said the new agents are built on Gemini 3.1 Pro and let developers combine web results with private data sources in a single research workflow. (blog.google) A 3D model here is not just a picture but an object you can rotate or adjust, like dragging a digital prototype on a screen. Google said users can change sliders or type in values such as gravity strength and initial velocity, then watch the simulation update inside the chat. (blog.google) Deep Research works like an automated research assistant that plans searches, reads sources, and assembles a cited report over several minutes. Google’s help pages say reports usually take 5 to 10 minutes, can use Google Search by default, and can also pull from sources such as Gmail, Drive, uploaded files, and NotebookLM notebooks. (support.google.com) Google split that research product into two modes. It said Deep Research is tuned for lower latency and lower cost, while Deep Research Max uses “extended test-time compute” to iterate more and produce more comprehensive reports for background jobs such as overnight due-diligence work. (blog.google) The release also shows how far Google’s product naming has moved since Gemini 2.5 Pro launched in March 2025 as its flagship “thinking model.” Google’s current developer and consumer materials now center Gemini 3.1 Pro for advanced research tasks, while Gemini 2.5 Pro remains documented in Vertex AI with a 1,048,576-token input limit and multimodal support for text, code, images, audio, and video. (blog.google) (docs.cloud.google.com) Google has been pushing that multimodal line for a year. In April 2025, it pitched Gemini 2.5 Pro to developers as a model for coding, agentic workflows, and reasoning over large code bases with a one million token context window. (developers.googleblog.com) The company is also tying those capabilities to subscription tiers. Google One’s pricing page says Google AI Pro includes higher access to Gemini 3.1 Pro and Deep Research, while Google AI Ultra includes the highest access to Gemini 3.1 Pro, Deep Research, Gemini Agent, and Deep Think. (one.google.com) Taken together, the update turns Gemini Pro into something closer to a workspace than a chatbot: one part visual sandbox, one part report generator. Google’s latest pitch is that the same family of models can sketch, simulate, search, and synthesize without leaving the chat window. (blog.google 1) (blog.google 2)