Jaque Silva | Sopa pictures | Lightrocket | Getty Images
Google On Wednesday, he published everyone Gimini 2.0 – his “most capable” model of artificial intelligence – everyone.
In December, the company approach Developers and reliable testers, as well as wrapping some features in Google products, but this is a “general edition”, according to Google.
Model sets include 2.0 flash, which is charged as “a model of a working horse, optimal for high -frequency, high -frequency tasks on a scale”; 2.0 Pro experimental, which is largely focused on coding performance; and 2.0 Flash-Lite, which Google charges as its “most payable model so far”.
Flash twins cost the developers 10 cents per million tokens for the entrances, pictures and video, while flash-lite, its more economical version, costs .75 of the Center for the same.
The continuation of the edition is part of the wider Google strategy for a big investment in “AI agents”, as AI arms of weapons are heated among technological giants and startups.
TargetAmazon, MicrosoftOpenai and Anthropic are also moving towards the agency AI or models that can complete complex multiple steps on behalf of the user, not that the user must carry them through each individual step.
“Over the past year, we invest in developing multiple agent models, which means that they can understand more about the world around you, think about more steps in advance and take actions on your behalf, with your supervision,” Google wrote in December in December blog blogAdding that twins 2.0 has “new progress in multimodality – such as the original image and audio output – and original tool use” and that the family of the model will allow us to build new agents of AI who bring us closer to our vision of the universal assistant. “
Anthropically, Amazon-If AI startup found by former research executives, is a key competitor in the race for the development of AI agents. In October startup said That his AI agents could use computers like people to complete complex tasks. The ability to use the ANTHROPICA computer allows its technology to interpret what is on the computer screen, select buttons, enter the text, move on to websites and perform tasks through any software and internet review in real time, is stated.
The tool can “use computers basically the same way as we do,” said Jared Kaplan, Chief Scientific Director of Anthropic, for CNBC in the then interview. He said he could do this with “dozens or even hundreds of steps”.
Openai posted a a similar tool Recently, introducing a feature called an operator that will automate tasks such as resting rest, filling in forms, making restaurant reservations and food ordering. IN Microsoft-Babard startup described the operator as “an agent who can go to the web to do tasks for you.”
Earlier this week, Openii announced another tool called deep research This allows AGTA to compose complex reports on research and analyze the questions and topics of the choice of users. In December, Google launched a similar tool of the same name – Deep Research – which acts as “assistant research, exploring complex topics and compiling a report on your behalf.”
CNBC reported for the first time in December To make Google a few features of AI in early 2025.
“You don’t always have to be the first in history, but you have to do well and really be the best in the class as a product,” said Sundar Pichai CEO at the time of the strategy at the time. “I think it’s 2025.”
