Mountain view, California: If you ever want a crisp summary of what you have achieved in a small window by Artificial Intelligence (AI), Beautiful pichaiChief Executive Officer, Google And Alphabet Keep it clearly. “More intelligence is available, for all, everywhere. And the world is responding, adopting AI much faster than ever … All this means that we are in a new phase of AI platform shift. Where decades of research research is now becoming a reality for people, businesses and communities around the world,” he said.
There are very few certainty in the world. It is the day after night. And there is Google’s annual I/o developer conferenceWhich sets the ball rolling for the company’s application and broad portfolio of services. Google Gemini 2.5 model, new generative AI models VO3 and imagene 4, AI filmmaking Tool Flow, Mithun’s frequent upgrade in individual pitch, AI finds more relevance in search, if you are ready to pay more for Google’s AI services, and then a universal AI is a universal AI accessory to create a universal AI services, which is a number of visions for the construction Is for people.
Ht was Detailed major changes Already prepared for Android this year, I/O on the first day of Keenote, with extensive measures to fight spamors and scammers. A little surprising, then, this attention changes correctly for AI conversation.
Gemini as a universal AI agent
Google, of course, is not alone in this conversation. The AI agents remain a continuous theme, some OpenAI, IBM, Anthropic and Microsoft, recently, have also made a case. Some call him “AI Agent” or “Agentic AI”, Google calls it a universal AI agent. Its key will have the ability to imitrse world knowledge, logic and natural environment, such as a human brain.
“Our recent updates for Gemini are important steps towards unlocking our vision for a universal AI assistant, one that is helpful in your everyday life, it is intelligent and understands the context you are in, and it can take action on your behalf in any device.
Hasbis understands it as an “AI who is intelligent, understands it
You are in reference, and it can take action on any device, action on its behalf and take action. Gemini model will provide foundations.
This project will be a culmination of merrinner, which “examines the future of human-agent interaction starting with browsers”, as well as project estra, video understanding, screen sharing and memory. Views now include a system of agents that can complete ten different tasks at a time. These tasks may include information, booking, buying things, and doing research on a subject, all can join in parallel.
Microsoft, in the build this week, widely expanded the launch of Windows AI Foundry as a foundation for the future of a native model reference protocol (MCP) and AI agents in Windows.
Anthropic introduced an open-source standard to MCP last year. It is also called “USB-C Port of AI”. Simplicity and comprehensive-spread support is important, as app developers can use MCP to enable their apps or agents to talk to other apps and services.
“We added native SDK support for model reference protocol (MCP) definitions at Mithun API for easy integration with open-source tools. We are also looking for ways to deploy MCP servers and other host tools, which makes it easy for you to create agent applications,” said “Tulca,” said, “Tulsi Doshi, Product Management Tulsi said.
Model updates, with long -term vision
Google is upgrading a significant upgrade to Gemini 2.5 flash and Gemini 2.5 Pro models. Add new logic capabilities with improvement in the Gemini 2.5 Pro model, deep think mode. Its specific focus on complex mathematics and coding works will be relevant to ‘Agentic AI’ vision for the march of Gemini.
Lighter Gemini 2.5 flash receives better logic, multimodality, code and long reference. For now, the update 2.5 flash is available as ‘experimental’ at AI Studio for Flash Developers, in vertex AI for enterprises, and Gemini app for all – its final release is in early June.
“Because we are defined the frontier with 2.5 Pro Depathink, we are taking extra time to evaluate more frontier safety and getting further input from security experts. As part of it, we are going to provide for reliable testers through Mithun API to get our feedback before making it widely available,” Core Kavukoglu, Core Kavukoglu, Core Kavukugoglu explained.
New creative generative ai mode
Google’s latest generic media models are now coming. Video generation models VO3 and image generation model imagene 4, get new capabilities. With new updates for the previous generation VEO 2 models, including the camera control, such as video generation elements such as camera movement or zoom, as well as better references from images of views, characters and objects for video generations.
“We are also expanding access to Leria 2, giving musicians more equipment to create music. Finally, we are inviting visual storytellers to try to flow, our new AI film manufacturing equipment. Using the most advanced model of Google deepmind, Flow lets you weave cinematic films with more refined controls for your story,”
The VO3 can now generate videos with audio, such as traffic noise in the background of the City Street scene, or even a dialogue between the characters, as well as better replication of real-world physics, lip-syncing and better understanding of signs.
Meanwhile, imagene 4, comes with the promise of recreation of better details.
“Imagene 4 has a significant clarity in excellent details in complex clothes, water drops, and animal fur, and both photorialistic and abstract styles. Imagene can create images in a range of 4 aspect ratio and up to 2K resolution – is also better for printing or presentations. It is also better in spelling and typography, which is easy to make its own.
Imagene 4 is now available in Gemini app, whisk, vertex AI and slides, VIDS, Docs in workspace. Availability of VO3, for now, is limited in Ultra subscribers in the US and Gemini app in flow.