Gemini 2.5 Flash continues to push the pareto frontier, so much intelligence packed into this model, can't wait for GA in a few weeks!
GENERATIVE AI
-
Veo 3 Enables Native Audio and Dialogue in Video Generation
By
–
Veo 3 has video gen with native audio – but also, it can do dialogue!
-

Google’s Imagen 4 Demo: AI-Generated Look Raises Quality Questions
By
–
Could Google have picked a better demo image for Imagen 4 that didn't look 100% AI generated?
-
Google Rolls Out Imagen 4 and Veo 3 AI Models
By
–
Imagen 4 and Veo 3 (with native audio), our SOTA image and video models, are rolling out right now and will be available to start in the @GeminiApp API soon : )
-
Google Search AI Mode Rolls Out to All US Users
By
–
AI mode in Google Search is starting to roll out to all users in the US today : )
— Logan Kilpatrick (@OfficialLoganK) 20 mai 2025
It's the search experience you know, reimagined for the AI era (including a new feature called "Deep Search")! pic.twitter.com/5njwWJWGD6AI mode in Google Search is starting to roll out to all users in the US today : ) It's the search experience you know, reimagined for the AI era (including a new feature called "Deep Search")!
-
Google Launches Project Mariner AI Agent Capabilities
By
–
Finally, we're so excited to bring Project Mariner capabilities to AI Mode in Search, Agent Mode in @GeminiApp, and as a standalone web app. This will allow the products you know and love to intelligently manage complex multi-step tasks on your behalf. https://t.co/H6IUpE0X5l pic.twitter.com/FPwpAw4eXd
— Google AI (@GoogleAI) 20 mai 2025Finally, we're so excited to bring Project Mariner capabilities to AI Mode in Search, Agent Mode in @GeminiApp
, and as a standalone web app. This will allow the products you know and love to intelligently manage complex multi-step tasks on your behalf. http://
labs.google/mariner -

Gemini Deep Think: Enhanced Reasoning Mode for 2.5 Pro
By
–
In the spirit of exploring the boundaries of Gemini's thinking capabilities, we're starting to test an enhanced reasoning mode of 2.5 Pro called Deep Think. It uses parallel thinking techniques to give Gemini time to consider multiple hypotheses before responding.
-
Gemini Diffusion: Faster Model Matches Advanced Coding Performance
By
–
Gemini Diffusion, our newest research model, is significantly faster than our fastest model so far AND matches its coding performance. By correcting errors as the model thinks, it is extremely fast for editing tasks like math and coding. pic.twitter.com/H811OrRebY
— Google AI (@GoogleAI) 20 mai 2025Gemini Diffusion, our newest research model, is significantly faster than our fastest model so far AND matches its coding performance. By correcting errors as the model thinks, it is extremely fast for editing tasks like math and coding.
-

Google Integrates LearnLM into Gemini 2.5 for Personalized Learning
By
–
With the help of experts in education, we have incorporated LearnLM into Gemini to make 2.5 the world's leading model for learning. Grounded in learning science principles, the model can adapt to individual learning styles to explain complex concepts in a clear way.
