Google Launches Gemini Windows App as DeepMind Faces Frontier Model Delays

Google has officially launched its native Gemini desktop application for Windows, bringing its artificial intelligence assistant directly to Windows 10 and 11 PCs. The release, which follows the macOS launch in April 2026, allows users to summon the AI assistant system-wide without keeping a web browser tab open, signaling Google's push to embed its AI ecosystem directly into desktop workflows.

While the desktop rollout marks a practical step forward for consumer accessibility, it arrives at a turbulent moment for Google's broader AI strategy. The tech giant is currently grappling with the delayed launch of its next-generation frontier model, Gemini 3.5 Pro, alongside a major leadership reorganization at Google DeepMind that has fueled discussions regarding the company's competitive standing against rivals like OpenAI and Anthropic.

The new Windows application is designed to run quietly in the background as a lightweight program, minimizing memory usage while remaining instantly accessible. Users can call up the Gemini overlay at any point by pressing the Alt + Space keyboard shortcut, which mirrors the Option + Space shortcut utilized in the macOS version. This overlay allows users to ask quick questions, draft text, or fact-check information without interrupting their active tasks or switching windows.

Beyond basic text queries, the desktop app integrates Google's proprietary media generation tools directly onto the PC desktop. Users can generate custom images using the Nano Banana model, while video generation is handled by Gemini Omni. Google has designed the application to support both x64 and ARM64 architectures on Windows 10 and 11, ensuring compatibility across a wide range of modern personal computers.

A core feature of the desktop application is its deep integration with Google Workspace services, including Gmail, Google Drive, Google Docs, and Google Calendar. By linking these accounts, users can instruct Gemini to perform complex, multi-step tasks, such as summarizing long email threads or compiling project summaries from documents stored in Drive. The app also utilizes saved account credentials to log users in automatically, streamlining the initial setup process.

However, the initial Windows release lacks several advanced capabilities that are currently exclusive to the macOS version. For instance, the ability to control Gemini using voice commands, summarize active file contents in Google Docs, and allow the AI to process information by directly sharing the user's screen are not yet available on Windows. Google has stated that these features will be gradually rolled out to Windows users in future updates.

The launch of the Windows desktop app and lighter-weight models like Gemini 3.6 Flash stands in stark contrast to the ongoing delays surrounding Google's premier AI model, Gemini 3.5 Pro. Originally announced by Alphabet CEO Sundar Pichai at a pre-Google I/O briefing in May with a promised June release, the model has yet to materialize. During that May briefing, Pichai stated, "We are also excited for 3.5 Pro. We are using it internally. It's showing great improvements. We are still testing and refining it, and it will roll out to everyone next month."

That rollout never occurred, and the delay has reportedly caused frustration among Google engineers, AI researchers, and managers who worry the company is losing its competitive edge. While Google stalls, its primary competitors have surged ahead with major releases. OpenAI recently launched its next-generation GPT-6 Astra model, and Anthropic introduced Fable 5.1, both of which have demonstrated advanced cyber coding capabilities and resolved complex mathematical research problems. Consequently, Google has slipped on public AI leaderboards, falling to 14th place overall on the Arena leaderboard and 25th place on the coding-specific leaderboard, trailing behind models from rivals and Chinese labs like Moonshot and DeepSeek.

The delays have coincided with a major structural shakeup within Google DeepMind. In August, Sundar Pichai announced that Demis Hassabis, the co-founder of DeepMind, would transition to the role of chairman of Google DeepMind and chief scientist of Alphabet. Senior Vice President Koray Kavukcuoglu, an AI researcher based at Google's headquarters in Mountain View, California, was appointed to lead the lab, effectively shifting the center of gravity away from DeepMind's historic London base.

This transition, which shocked many employees, occurred alongside the departure of legendary AI researcher Jeff Dean, who left Google to start a new venture. Analysts have questioned whether the reorganization represents a consolidation of power in Mountain View or a response to a challenging year marked by delayed models, talent exits, and internal friction over military defense contracts. Google, however, has publicly characterized the leadership changes as a promotion for Hassabis.

Google has indicated that it is actively testing Gemini 3.5 Pro alongside an upgraded Flash model with external partners, while also engaging with the United States government on model testing and safety frameworks. During the Alphabet second-quarter earnings call, Pichai sought to reassure investors by revealing that the company is already training its next-generation model, Gemini 4, with plans to maintain a rapid, near-monthly cadence of model updates. Pichai noted, "We are now training Gemini 4, and we're being very ambitious with it."

For desktop users, Google plans to introduce more native desktop capabilities to the Windows app over the coming months to bridge the feature gap with macOS. The installer is currently available for direct download from Google's official website, allowing Windows 10 and 11 users to integrate the assistant into their daily workflows immediately.