Google has officially introduced Gemini 3, its newest and most advanced AI model suite yet, designed to replace the Gemini 2.5 version that debuted last May at Google I/O.
Gemini 3 is now available in AI Mode on Google Search for Google AI Pro and AI Ultra subscribers, as well as through the Gemini app accessible to all users. Sundar Pichai, Google’s CEO, emphasizes that this model is "much better at understanding the context and intent behind your queries, delivering what you need with less effort."
Boasting the title of "the best model worldwide for multimodal comprehension," Gemini 3 introduces richer visual outputs and more interactive experiences. The rollout begins with Gemini 3 Pro, currently in preview, with Gemini 3 Deep Think following soon after. The latter is undergoing rigorous safety testing before becoming available to Google AI Ultra subscribers.
According to Google, Gemini 3 Pro "significantly outperforms" Gemini 2.5 Pro across all major AI benchmarks. It offers nuanced and direct responses—prioritizing authentic insight over flattery, ensuring users receive honest and meaningful feedback rather than just agreeable answers. Gemini 3 Deep Think surpasses Gemini 3 Pro in performance metrics but takes additional time to generate its responses due to deeper analysis.
Built to help users learn more effectively, Gemini 3 integrates information from multiple modalities including text, images, video, audio, and code, enabling seamless synthesis across diverse content types.
Security enhancements mark Gemini 3 as Google's most robust AI model so far. It has "undergone the most exhaustive safety evaluations in Google AI’s history," featuring reduced tendency for sycophantic responses, improved defenses against prompt injections, and stronger safeguards against potential cyber misuse.
