Google launches Gemini 3, a new era of intelligent multimodal AI

What happened

Google launched Gemini 3, its most advanced multimodal AI model, on November 18, 2025, integrating it across Search, the Gemini app, AI Studio, Vertex AI, and other platforms. Gemini 3 combines state-of-the-art reasoning, multimodal understanding, and agentic coding to handle complex tasks involving text, code, and rich media. The release includes Gemini 3 Pro available immediately and Gemini 3 Deep Think, a higher-intensity reasoning mode, to be rolled out soon to premium subscribers.

Why it matters

• Gemini 3 unifies multiple AI capabilities into a single model, enabling seamless analysis of text, images, videos, and documents in one request.
• It enhances Google Search and enterprise applications with advanced reasoning and interactive AI experiences, improving problem-solving and planning tasks.
• The model’s multimodal and agentic features support complex workflows like financial analysis, supply-chain planning, and media-heavy analytics, expanding AI’s practical utility.

Key facts

1. 1. Gemini 3 was officially launched on November 18, 2025.
2. 2. Gemini 3 Pro is available immediately across Google Search, the Gemini app, AI Studio, Vertex AI, and Google Antigravity IDE.
3. 3. Gemini 3 Deep Think mode, offering enhanced reasoning, will be available in the coming weeks to Google AI Ultra subscribers after safety testing.
4. 4. Gemini 3 can process combined inputs of text, media, and documents in a single request, supporting long PDFs, screenshots, and video snippets.
5. 5. Gemini 3 demonstrated top-tier performance on benchmarks like Humanity’s Last Exam (41.0%) and ARC-AGI (45.1% with code execution).

Key Voices

• Quoc Le, Google AI researcher, highlighted Gemini 3 Deep Think’s superior reasoning, citing its gold medal-level wins at international math and programming contests.
• Bob Bradley, VP at Geotab, praised Gemini 3’s multimodal capabilities in enterprise applications, noting its superior performance in real-world audio and vision tasks.
• Google’s blog emphasized Gemini 3’s ability to grasp nuance and context with less prompting, marking a significant evolution in AI understanding and interactivity.
• Not stated in the article.