Gemini News and Updates (2026)

The familiar Gemini interface is going away. Google began rolling out a ground up redesign on iOS that replaces the static homepage with a pulsating gradient, centers the greeting prompt under the spark icon, and collapses tools like Deep Research, Canvas, and Guided Learning into one organized list. The prompt box is now pill shaped. The model picker sits in a top left dropdown. On iOS, Apple's Liquid Glass effects give everything a polished, layered feel. It is the most significant UI change since Gemini launched.

General Motors will push Gemini to model year 2022 and newer Cadillac, Chevrolet, Buick, and GMC vehicles with Google built in. Around 4 million U.S. cars are eligible for the over the air update, one of the largest AI assistant deployments in the automotive industry to date.

After a preview period that saw developers build everything from e commerce discovery engines to video analysis tools, Google moved Gemini Embedding 2 to GA across the Gemini API and Vertex AI. The model maps text, image, video, and audio into a unified embedding space, which simplifies multimodal search and retrieval pipelines considerably.

Two new Deep Research models landed in the API: a streaming variant optimized for real time client UIs and a max variant built for comprehensive autonomous gathering. Both support collaborative planning, visualization output, and File Search. The headline for developers: native Model Context Protocol server integration, making Gemini the first major AI provider to bake MCP into its research agent layer.

This is the update that makes Gemini feel less like a chatbot and more like a coworker. Workspace Intelligence gives the assistant persistent context from Gmail, Calendar, Chat, and Drive, so it no longer needs you to re explain your projects in every conversation. Scheduled actions let you set up recurring tasks (morning calendar summaries, weekly project idea brainstorms, Friday event roundups) that run without prompting. A new Productivity Planner Gem ties it all together.

A packed monthly drop. The Gemini app launched natively on macOS for the first time. NotebookLM integration arrived in the app as Notebooks, letting you manage research projects alongside conversations. Lyria 3 Pro opened up 3 minute music track generation for free. Personal Intelligence now uses your connected Google apps to generate images that actually reflect your life and interests. And a new visualization feature turns complex concepts into interactive explainers directly in chat.

Chrome users on Windows, macOS, and Chromebook Plus can now open Gemini in a side panel that persists while browsing. The update includes Nano Banana image editing without leaving the active tab. More interesting: an auto browse preview that lets Gemini navigate websites and complete tasks on your behalf while you stay in control. Available to Google AI Pro and Ultra subscribers in the U.S.

A small but useful addition: Gemini can now generate PDFs, Word docs, Excel spreadsheets, Google Docs, Sheets, and Slides directly inside a conversation. Instead of copying output into another app and reformatting, you get a downloadable file you can save locally or push to Drive. Available globally to all Gemini app users.

The branding cleanup that was long overdue. Google retired both the Gemini Advanced name and the Google One AI Premium name, replacing them with a clean three tier stack. Google AI Plus ($7.99 per month) is the entry level paid tier. Google AI Pro ($19.99 per month) carries everything Gemini Advanced had plus Veo 3.1 video generation and NotebookLM Plus. Google AI Ultra ($249.99 per month) targets power users with Deep Think reasoning, 30 TB storage, $100 in monthly Google Cloud credits, YouTube Premium, and the highest rate limits across every Gemini surface.

The original gemini-3-pro-preview endpoint was deprecated, with all traffic redirected to gemini-3.1-pro-preview automatically. No developer action required, but Google recommended updating API calls to the 3.1 string for long term compatibility.

Three months after Gemini 3 Pro rewrote the leaderboards, Google shipped its first major iteration. Gemini 3.1 Pro sharpened complex problem solving across the board. Deep Think, a specialized reasoning mode for science and engineering, expanded to Google AI Ultra subscribers. Nano Banana 2 (built on 3.1 Flash Image) brought noticeably higher fidelity image generation with accurate text rendering in any language. Quietly, Gemini also started linking directly to scientific papers in its responses, a significant move for research credibility.

After debuting in Indonesia in September 2025, the entry level paid tier came to the U.S. at $7.99 per month (50% off the first two months). It includes a 128K context window, 200 monthly AI credits for creative tools, expanded NotebookLM access, and 200 GB of cloud storage. A sensible on ramp for users who outgrow the free tier but do not need Pro level access.

Apple confirmed plans to integrate the Gemini model into an upcoming version of Siri. Details on scope and timeline were thin, but the announcement itself was the story: it could put Gemini on hundreds of millions of iPhones and iPads, giving Google an AI distribution channel that no amount of marketing spend could replicate.

At CES 2026, Google showed Gemini running on Google TV with natural language content discovery, series recaps, educational deep dives, and voice controlled settings. Instead of digging through menus, you tell Gemini the screen is too dim and it adjusts. First rollout targets select TCL televisions running Android TV OS 14 or higher.

The model that most free users interact with got a generational upgrade. Gemini 3 Flash replaced 2.5 Flash as the default, bringing what Google called PhD level reasoning to the fast tier. Multimodal understanding across images, audio, and text jumped significantly. Users pick between Fast (3 Flash for quick answers) and Thinking (3 Pro for complex problems) in a simple toggle.

A preview of the Gemini Deep Research Agent went live in the API. It autonomously plans multi step research tasks, decomposes queries, evaluates sources, and synthesizes results into structured reports. No manual intervention needed between steps. This early version laid the groundwork for the more capable Deep Research agents that shipped in April 2026.

Built on the Gemini 3 Pro Image architecture, Nano Banana Pro improved text rendering and world knowledge over the original Nano Banana. Given that the original had already become the most popular AI image generator within weeks of its August launch, the Pro variant set a high bar for competitors in the image generation space.

After incremental 2.5 updates throughout mid 2025, Google skipped straight to a generational leap. Gemini 3 Pro debuted at 1,501 Elo on the LMArena leaderboard, becoming the first model to break 1,500. The benchmark number alone was not the story. It shipped alongside Deep Think for extended reasoning, Antigravity (a new agentic IDE where AI manages your editor, browser, and terminal), and same day availability across Workspace, Vertex AI, and Google AI Studio. Multimodal understanding, agentic coding, tool use, and long context processing all improved substantially over 2.5 Pro.

Two weeks after appearing anonymously on the Arena evaluation platform, Google revealed Nano Banana's identity: Gemini 2.5 Flash Image. The name stuck. Within weeks it became the most widely used AI image generator, beating established competitors on speed and quality. The codename, a nod to the nicknames of Google DeepMind Product Manager Naina Raisinghani, became part of the product's official branding.

The full 2.5 model family reached general availability: Pro for advanced reasoning ($1.25 per million input tokens, $10.00 per million output tokens), Flash for balanced performance, and the new Flash Lite for cost sensitive high volume workloads. All three shipped with a 1 million token context window. Enterprise customers could access them through both Vertex AI and Google AI Studio, marking the point where 2.5 moved from preview curiosity to production workhorse.