Jaque Silva | Sopa images | Lightrocket | Getty images
Google Wednesday, published Gemini 2.0 – His sequence of artificial “most competent” artificial “models to date – to everyone.
In December, the company gave access to developers and trusted testers, as well as to pack certain features in Google Products, but it is a “general version”, according to Google.
The suite of models includes 2.0 flash, which is presented as a “battle horse model, optimal for high frequency tasks at high volume”; 2.0 Experimental pro, which is largely focused on coding performance; And 2.0 flash-lite, which Google invoices as “its most profitable model to date”.
Gemini Flash costs developers 10 cents per million tokens for text, image and video entrances, while Flash-Lite, its more profitable version, costs 0.75 central for the same.
Continuous outings are part of a broader strategy for investment massively in “AI agents”, because the AI’s arms race heralds among the giants and the technology startups.
MetaAmazon, MicrosoftOPENAI and Anthropic also go to an agentic AI, or models that can perform complex tasks in several steps in the name of a user, rather than a user in front of him browsing each individual stage.
“In the past year, we have invested in the development of more agent models, which means that they can better understand the world around you, think several steps and take measures on your behalf, with your supervision”, wrote Google in December blogAdding that Gemini 2.0 has “new advances in multimodality – such as native image and audio output – and the use of native tools” and that the family of models “will allow us to build new AI agents that bring us closer to our vision of a universal assistant. “
Anthropic, the IA startup supported by Amazon founded by the ex-operators of research-operai, is a key competitor of the race to develop agents of the AI. In October, the startup said that its AI agents were able to use computers like humans to perform complex tasks. The capacity of anthropic’s computer allows its technology to interpret what is on a computer screen, select buttons, enter text, navigate websites and d ‘Performing tasks via all software and internet browsing in real time, said start -up.
The tool can “use computers in the same way as us,” said Jared Kaplan, Director of Sciences of Anthropic, in CNBC in an interview at the time. He said it can do tasks with “tens or even hundreds of steps”.
OPENAI recently published a similar tool, introducing a functionality entitled Operator which will automate tasks such as vacation planning, filling of forms, restaurant reservation and grocery command. The operator described Microsoft’s back starting as “an agent who can go to the web to perform tasks for you”.
Earlier this week, Openai announced another tool called Deep Research which allows an AI agent to compile complex research reports and analyze the questions and subjects of the user’s choice. Google launched in December a similar tool for the same name – Deep Research – which acts as a “research assistant, exploring complex subjects and compiling reports on your behalf”.
CNBC reported in December for the first time that Google would present several AI features in early 2025.
“In history, you don’t always need to be the first, but you have to do well and be really the best in class as a product,” said CEO Sundar Pichai at a meeting of strategy at the time. “I think that’s what 2025.”
