Google published a sequence of Gemini release notes highlighting major product updates and model rollouts. The top entry, dated 2025.12.04, announces Gemini 3 Deep Think, described as Gemini’s most advanced reasoning mode. Google Artificial Intelligence Ultra subscribers can access Gemini 3 Deep Think by selecting Deep Think in the prompt bar and choosing ‘Thinking’ in the model dropdown; the company says responses are typically ready in a few minutes. Deep Think uses iterative rounds of reasoning and multiple parallel hypotheses to tackle complex problems in math, science and logic, aiming to improve reasoning capability over prior models.
Earlier entries explain broader availability of Gemini 3 Pro (2025.11.18) and platform integrations. Gemini 3 Pro rolled out to everyone in the Gemini app with higher limits for Google Artificial Intelligence Plus, Pro and Ultra plans; users are advised to pick ‘Thinking’ in the model drop-down to try it. On 2025.11.20, Gemini replaced Google Assistant in Android Auto for a more conversational Artificial Intelligence experience that connects to user apps for hands-free tasks. The release notes also introduce experimental Gemini Labs features: visual layout creates multimodal, interactive responses and dynamic view uses agentic coding to build custom interfaces on the fly. Both are being rolled out to subsets of users for feedback and iteration.
The notes recap subscription and creative tool changes that appeared across the year. Student promotions and subscription rebrands are detailed: eligible university students aged 18 and older may sign up for a one-year free upgrade to Google Artificial Intelligence Pro by 9 December 2025, which includes image uploads and 2 TB of storage. Google Artificial Intelligence Ultra subscribers receive priority access to advanced models and features such as Veo 3 for 8-second videos with sound and a 1M token context window. Document upload and context-window details are explained: a 32K context window (around 50 pages of text) is available to general users, while Gemini Advanced users retain a 1M token context window (around 1,500 pages of text).
The release notes emphasize rapid iteration and experimental rollouts: upgrades to 2.5 Pro and 2.5 Flash, Canvas and Deep Research integrations, Gems sharing, scheduled actions, temporary chat options and other features are being tested with subsets of users. Google repeatedly frames these launches as experimental in some tiers, asking for user feedback to guide wider releases while noting that experimental models can behave unexpectedly.
