Highlights
- Google launched Gemini 3, its most advanced AI model yet that outperforms GPT-5.1 and Claude 4.5 Sonnet.
- The lineup includes Gemini 3 Pro available via Gemini apps, Search, AI Studio, Vertex AI, and Gemini 3 Deep Think is limited to safety testers, later for AI Ultra users.
- Google introduced Antigravity, an AI-powered development platform.
Caption – CEO Sundar Pichai introduces Gemini 3. (Image credit – Google)
Google has officially unveiled the Gemini 3 family of AI models, calling it the company’s most powerful and intelligent AI system to date. According to the tech giant, Gemini 3 outperforms its previous models as well as OpenAI’s GPT-5.1 across all major benchmark tests. The new generation brings major advancements in reasoning, conversational ability, coding, mathematics, and agentic capabilities. It positions Gemini 3 as a major leap toward more complex, autonomous AI experiences.
In a blog post, CEO Sundar Pichai described the new model as “our most intelligent model.” In an X post introducing the model, Pichai wrote, “It’s the best model in the world for multimodal understanding, and our most powerful agentic + vibe coding model yet. Gemini 3 can bring any idea to life, quickly grasping context and intent so you can get what you need with less prompting.”
Elon Musk and Sam Altman React to Gemini 3
The launch quickly drew responses from prominent AI leaders. Elon Musk, CEO of xAI and one of Google’s biggest rivals in the AI space, reacted on X shortly after Pichai announced “Geminiii,” seemingly a playful nod to the Roman numeral for three. Musk replied with a simple “Congrats.”
OpenAI CEO Sam Altman also acknowledged the development. Posting on X, he wrote, “Congrats to Google on Gemini 3! Looks like a great model.”
Gemini 3 – Features, Models, and Availability
Google announced the Gemini 3 lineup through an official blog post after several cryptic hints from executives on X. The series includes two models, Gemini 3 Pro and Gemini 3 Deep Think.
Google explains, “Gemini 3 Pro can bring any idea to life with its state-of-the-art reasoning and multimodal capabilities. It significantly outperforms 2.5 Pro on every major AI benchmark. It tops the LMArena Leaderboard with a breakthrough score of 1501 Elo. It demonstrates PhD-level reasoning with top scores on Humanity’s Last Exam (37.5% without the usage of any tools) and GPQA Diamond (91.9%). It also sets a new standard for frontier models in mathematics, achieving a new state-of-the-art of 23.4% on MathArena Apex.”
Gemini 3 Deep Think, on the other hand, “pushes the boundaries of intelligence even further, delivering a step-change in Gemini 3’s reasoning and multimodal understanding capabilities to help you solve even more complex problems.”
The rollout is happening at scale, meaning users will encounter Gemini 3’s capabilities across the Gemini apps and website, AI Mode in Search, AI Studio, and Vertex AI for developers. Google also introduced a new development environment called Google Antigravity to support agentic AI workflows.
Reliance Jio confirmed that all users subscribed to the free Google AI Pro plan will gain access to Gemini 3 Pro.
However, availability varies between the two models. Gemini 3 Pro is the most advanced variant accessible to users and is launching in preview and will be integrated into Google’s suite of products. Meanwhile, Gemini 3 Deep Think is designed for enhanced reasoning and is currently restricted to safety testers and will later be made available exclusively to Google AI Ultra subscribers.
Performance – Outshining GPT-5.1 and Claude 4.5 Sonnet
Google shared early benchmark data for Gemini 3 Pro based on internal evaluations. The results show the model outperforming both its predecessor and OpenAI’s GPT-5.1 on every major benchmark. It also beats Anthropic’s Claude 4.5 Sonnet in all but two tests, AIME 2025 (math) and SWE-Bench (coding).
Caption – Gemini 3 across a range of key AI benchmarks. (Image from Google Blog)
On Humanity’s Last Exam, an advanced academic reasoning benchmark considered extremely challenging, Gemini 3 Pro scored 37.5 per cent without tools. This surpasses the previous high score of 26.5 per cent set by GPT-5.1. On the ARC-AGI-2 visual reasoning benchmark, a key measure of artificial general intelligence (AGI), Gemini 3 Pro achieved 31.1 per cent, significantly higher than GPT-5.1’s 17.6 per cent and Gemini 2.5 Pro’s 4.9 per cent.
Google also claims substantial progress in coding performance, especially in frontend development. Gemini 3 Pro can now generate complete functional websites, apps, and scalable vector graphics (SVGs).
For Gemini 3 Deep Think, Google reports even higher figures. It scores 41.0 per cent on Humanity’s Last Exam and 45.1 per cent on ARC-AGI-2. However, this model is not yet available to users.
Google Antigravity – A New Agentic Development Platform
Alongside Gemini 3, Google introduced Antigravity. It is an agentic development platform designed for developers to work alongside AI agents. It acts as an AI-driven integrated development environment (IDE) with access to Gemini 3’s reasoning abilities, tool use, and agentic coding features. Agents operate in isolated workspaces equipped with an editor, terminal, and browser.
These agents can autonomously plan, write, execute, and verify code, allowing them to handle complex end-to-end software tasks. Antigravity also includes the latest Gemini 2.5 Computer Use model for browser automation and Nano Banana, in addition to Gemini 3 Pro.
FAQs
Q1. What is Gemini 3 and how does it compare to other AI models?
Answer. Gemini 3 is Google’s most advanced AI model to date, outperforming GPT-5.1 and Claude 4.5 Sonnet across major benchmarks in reasoning, coding, and AGI.
Q2. What models are included in the Gemini 3 lineup and who can access them?
Answer. The lineup includes Gemini 3 Pro (available via Gemini apps, Search, AI Studio, Vertex AI) and Gemini 3 Deep Think, which is currently limited to safety testers and will later be available to Google AI Ultra subscribers.
Q3. What is Google Antigravity and how does it relate to Gemini 3?
Answer. Google Antigravity is a new agentic development platform that lets developers collaborate with AI agents to autonomously plan, write, and test code using Gemini 3’s reasoning and tool-use capabilities.