News
Google Pushes Gemini 3 Pro Into Search And The Gemini App
The new model lands a few weeks before Gemini 2’s first birthday and brings a task-running agent to the Gemini app.
Google has pushed Gemini 3 Pro across its consumer and developer platforms, dropping the model into the Gemini app, Google Search’s AI Mode and other products across the company’s cloud stack in a single sweep.
In the app, the model sits behind the “Thinking” option and is open to all users, with higher limits for paid tiers. Answers are reported to be tighter and better structured. More importantly, it unlocks Gemini Agent, a task-runner built on last year’s Project Mariner. Once granted access to Google apps, it can handle chores directly — managing email instead of just suggesting how.
Search has also received a major upgrade. Gemini 3 Pro goes live first for AI Pro and Ultra subscribers inside AI Mode and will also power AI Overviews for the hardest queries. A new routing system (due soon), will decide when a question is sent to the upgraded model. For now, subscribers can trigger it by selecting “Thinking” in AI Mode.
Google argues the model strengthens the “fan-out” technique behind AI Mode by running broader, deeper searches and surfacing material earlier versions overlooked. Better multimodal analysis should also produce more functional answers; researching personal loans, for example, could return an embedded calculator rather than a wall of text.
The company is leaning on benchmark gains to anchor the launch. On benchmarking platform Humanity’s Last Exam, the model hit 37.5 percent accuracy, topping Grok 4 by 12.1 points without using web search. It also leads the LMArena board with 1,501 points.
Also Read: IBM Unveils Nighthawk And Loon Quantum Chips
Developers get the same upgrades too: Gemini 3 Pro is now in the Gemini API, AI Studio and Vertex AI. Google is also adding Antigravity, an autonomous coding tool that writes software while generating its own subtasks and progress notes. Gemini 3 Deep Think, a higher-end reasoning mode, will reach safety testers first before moving to AI Ultra users.
For MENA firms already prototyping with Google models, the rollout offers quicker access to production-grade assistants and coding agents as competition between global AI platforms tightens.