Google unveils "Thinking" mode and "Vibe Coding" capabilities in a massive update designed to overtake OpenAI’s GPT-5.1.
The battle for artificial general intelligence (AGI) entered a frantic new chapter on Tuesday as Google unveiled Gemini 3, its most potent AI model to date. The launch, arriving just seven months after its predecessor, signals Google’s aggressive push to reclaim dominance from rivals OpenAI and Anthropic in an industry now moving at breakneck speed.
Described by CEO Sundar Pichai as the "best model in the world for multimodal understanding," Gemini 3 is designed to not just process information but to "read the room," offering a significant leap in reasoning capabilities that the company claims surpasses OpenAI’s recently released GPT-5.1.
"Gemini 3 can bring any idea to life, quickly grasping context and intent so you can get what you need with less prompting," Pichai said in a statement accompanying the launch.
In the official announcement, Pichai stated: "It’s amazing to think that in just two years, AI has evolved from simply reading text and images to reading the room."
The "Thinking" Model
The centerpiece of the update is a new "Thinking" mode, accessible immediately to subscribers of Google's AI Premium and Ultra plans. Similar to OpenAI's reasoning models, this feature allows Gemini 3 to "pause" and deliberate before responding to complex queries—a process Google claims effectively eliminates the "hallucinations" that plagued earlier generations of Large Language Models (LLMs).
According to Demis Hassabis, CEO of Google DeepMind, the model’s responses have been engineered to be smart, concise, and direct, trading clichés and flattery for genuine insight—telling you what you need to hear, not just what you want to hear.
This focus on depth is backed by impressive benchmark claims. Google reports that Gemini 3 has already topped the LMArena Leaderboard with a record-breaking score of 1501 Elo and secured the number one spot on Humanity’s Last Exam (37.5%), a benchmark designed to test expert-level reasoning across the humanities and sciences.
"Vibe Coding" and Agentic Futures
While the consumer features are flashy, the most significant updates may be for developers. Google introduced a concept it calls "vibe coding," where the model understands the aesthetic and functional "vibe" of a coding project with minimal instruction.
To support this, Google launched Antigravity, a new agentic development platform. Unlike traditional coding assistants that suggest lines of code, Antigravity allows developers to command AI agents to build entire software modules autonomously.
Koray Kavukcuoglu, Chief Technology Officer at Google DeepMind, noted that the model can now "consume entire codebases" thanks to a massive 1-million-token context window, allowing it to refactor legacy systems or generate high-fidelity prototypes in seconds.
Generative UI: The End of Static Search?
The update also brings a radical change to how users interact with Google Search. A new feature dubbed "Generative UI" allows Gemini 3 to code dynamic, interactive interfaces on the fly.
Instead of a standard list of blue links or a text summary, a user asking to "plan a 3-day trip to Rome" might see a bespoke, magazine-style travel itinerary with interactive maps and booking modules created specifically for that query.
"We believe that the best way to answer a question isn't always a text-based response," Google noted in its official blog post. "By creating dynamic, interactive interfaces on the fly, we can help you explore information in a way that is more intuitive."
Gemini 3’s Availability
Gemini 3 is rolling out starting today.
- For Consumers: Available in the Gemini app and "AI Mode" in Google Search.
- For Developers: Accessible via Google AI Studio and Vertex AI.
- Coming Soon: A research-focused variant, Gemini 3 Deep Think, is currently in safety testing and will be released to "Ultra" subscribers in the coming weeks.
As the dust settles on this latest salvo, the question remains: Will Gemini 3 be enough to solidify Google's lead, or will OpenAI’s next move shift the ground once again? For now, the "Gemini Era" that Pichai promised two years ago seems to be going somewhere.
Check out our social post
