P
Back to articles
News4 min read

Gemini 3.5: Google's frontier intelligence model with action capabilities

Google unveiled Gemini 3.5 at I/O 2026, a frontier AI model with native action capabilities that can navigate websites, fill forms, execute code, and manage workflows. The model powers new AI Mode in Search, deeply integrated Workspace agents, and revamped Google Finance. Google describes it as 'frontier intelligence with action,' marking a shift from conversational to agentic AI.

Source: Google AI Blog
Gemini 3.5: Google's frontier intelligence model with action capabilities

Gemini 3.5: Google's Frontier Intelligence Model With Action Capabilities

Google unveiled Gemini 3.5 at its annual I/O 2026 developer conference, introducing what the company describes as its most advanced frontier intelligence model yet — one that features native action capabilities that mark a new era of agentic AI. The announcement was the centerpiece of a conference that Google framed as "Welcome to the agentic Gemini era."

Gemini 3.5 represents a generational leap over its predecessor, Gemini 2.5 and the earlier Gemini 3.0 series. While previous models excelled at understanding and generating content, Gemini 3.5 is designed to act — executing multi-step tasks, interacting with software tools, and taking actions in digital environments autonomously under human supervision.

Native Action Capabilities

The defining feature of Gemini 3.5 is its ability to perform actions, not just generate responses. The model can navigate websites, fill forms, control software applications, execute code, and manage complex workflows — all through natural language instructions. This represents a fundamental shift from conversational AI to agentic AI, where the model becomes an active participant in getting things done.

During the I/O keynote, Google demonstrated Gemini 3.5 performing tasks like researching and booking travel, managing email workflows, analyzing spreadsheets with multi-step operations, and even building simple applications from natural language descriptions. The model demonstrated continuity across tasks, remembering context and maintaining state throughout extended interactions.

New AI-First Products

Alongside Gemini 3.5, Google announced a wave of new products and features. AI Mode in Google Search represents "a new era for AI Search," integrating Gemini's reasoning capabilities directly into the search experience. AI Mode can handle complex, multi-part queries that traditional search struggles with — comparing products across multiple dimensions, synthesizing information from multiple sources, and presenting results conversationally.

Google Workspace received deep Gemini integration, with AI agents that can draft emails, summarize documents, create presentations, and manage calendar conflicts. Google Finance was revamped with AI-powered analysis, expanding to European markets. A new AI-powered Google Beam feature aims to improve group meeting experiences.

Model Architecture and Availability

Gemini 3.5 is available across Google's AI offerings, including Gemini Advanced subscriptions and the newly revamped AI subscriptions tier. The model powers Google's agentic experiences across Search, Workspace, and developer platforms. Google highlighted that Gemini 3.5 represents "frontier intelligence with action" — a model that doesn't just think, but does.

The I/O 2026 conference featured 100 product announcements, but Gemini 3.5 was the clear headliner — signaling Google's commitment to leading the transition from AI that responds to AI that accomplishes.

Related Articles

Fine-tuning GPT-2 from human preferences
News8 min

Fine-tuning GPT-2 from human preferences

In 2019, OpenAI researchers led by Daniel M. Ziegler published 'Fine-Tuning Language Models from Human Preferences,' the foundational paper that introduced reinforcement learning from human feedback (RLHF) for natural language tasks. Using GPT-2 as the base model, the team demonstrated that human preference judgments — collected through pairwise comparisons — could effectively fine-tune language models across four tasks: stylistic continuation with positive sentiment, physically descriptive continuation, and summarization on the TL;DR and CNN/Daily Mail datasets. The paper's most influential finding was that models trained with 60,000 human comparisons learned to copy whole sentences verbatim from source text, revealing how labeler heuristics can shape model behavior — a lesson that remains central to modern alignment research.

LLMs & Chatbots
Multimodal neurons in artificial neural networks
News5 min

Multimodal neurons in artificial neural networks

OpenAI discovered multimodal neurons in CLIP that respond to the same concept across visual, textual, and symbolic representations, revealing how AI models develop abstract conceptual understanding and offering new insights into both model capabilities and biases.

LLMs & Chatbots
Anthropic launches Cowork — a Claude Desktop agent that works in your files, no coding required
Review4 min

Anthropic launches Cowork — a Claude Desktop agent that works in your files, no coding required

Anthropic launched Cowork, a Claude Desktop agent for macOS that operates on user files through natural language — no coding required. Built on Claude Code's architecture but designed for knowledge workers, it can generate expense reports, organize documents, and synthesize notes. Available on Claude Pro ($20/month), it competes with OpenAI's desktop agents and Microsoft Copilot in the AI productivity space.

LLMs & Chatbots