Onome Amuge
Google has launched Gemini 3, its most advanced AI model yet, marking a shift from conversational tools toward autonomous task execution amid intensifying competition in next-generation artificial intelligence.
The launch, announced on Tuesday, introduces a family of models led by Gemini 3 Pro, which Google says brings “PhD-level reasoning” and unprecedented multi-modal understanding to everything from search and software development to personal productivity. It marks one of Google’s most aggressive attempts yet to reassert its dominance in AI amid fierce competition from OpenAI, Anthropic and a fast-expanding cohort of frontier model developers.
“This new chapter is about pushing the frontiers of intelligence, agents and personalization,” said Sundar Pichai, CEO of Google and Alphabet.
Gemini 3 Pro sets new benchmarks on complex reasoning tasks, scoring 91.9 per cent on GPQA Diamond, a test typically associated with graduate-level problem solving. But Google’s pitch goes beyond benchmark supremacy. The model is designed to interpret handwritten notes, understand long videos, manipulate complex data sets and carry out multi-step tasks with minimal prompting.
More striking is the company’s repeated emphasis on agentic capability, the ability for AI systems not just to produce text, but to navigate context, break goals into sub-tasks and execute actions across applications. In concrete terms, this means the AI can prioritise inboxes, draft structured responses, analyse financial scenarios, and build interactive simulations, rather than merely explain how to do those tasks.
Search transforms into a learning environment
The launch of Gemini 3 signals a major rethink for Google Search, long the company’s flagship product. In its new AI Mode, Search moves beyond links and summaries, generating interactive, real-time tools customized to each query. Physics students can manipulate a simulated three-body system, while prospective homebuyers receive bespoke mortgage calculators that adjust automatically to interest rates and repayment terms.
The company also points to changing global behaviour as evidence that demand for more capable AI systems is accelerating. In Nigeria, one of Google’s fastest-growing markets, search interest in “AI and studying” has risen over 200 per cent, while queries on “how to learn AI” have increased 160 per cent in the past year. Rising digital literacy and widespread mobile connectivity are driving the trend, reflecting broader shifts across emerging markets where AI is becoming a key tool for education, entrepreneurship and workforce mobility.
From assistance to autonomy: the new Gemini Agent
One of the most radical components of the rollout is the introduction of the Gemini Agent, an experimental personal assistant currently available to Google AI Ultra subscribers. Unlike earlier assistants, the Gemini Agent is designed to take real actions across connected applications.
A command such as “organise my inbox” triggers a sequence in which the agent reviews unread emails, identifies urgent tasks, drafts replies and groups messages into actionable categories. A request to “book a mid-size SUV for my trip” prompts the agent to extract flight times from email confirmations, compare rental options, and prepare a booking workflow, all while requiring final user approval.
To cement its position with developers, Google introduced Google Antigravity, a platform that turns AI into something closer to a software-building collaborator. Rather than producing snippets of code, Antigravity is designed to plan and execute multi-layered development tasks including debugging, building features, rewriting modules and coordinating multiple agents.
Google describes it as “mission control” for agentic systems, providing verifiable artefacts to ensure trust. The move positions Google to compete not only in consumer AI but in the emerging market for autonomous software engineering tools.
Safety and availability
Google says Gemini 3 has undergone its most comprehensive safety review to date, though the company offered limited detail about evaluation protocol or independent oversight. Gemini 3 Pro is available immediately through the Gemini app and through AI Mode in Search for Pro and Ultra subscribers.
