While OpenAI prepares its next offensive between its Orion projects and a future GPT-5 which is long overdue, Google does not give up an inch of ground and is preparing to unveil its new evolution Gemini 2.0.
It is predicted to have greater processing capacity and perhaps to have a function that will allow taking control of a computer interface via an agent Jarvis capable of managing the Chrome web browser (from Google) as competitor Anthropic already wants to offer with the function Compute Use usable with the Claude 3.5 Sonnet AI model.
Imminent launch for Gemini 2.0?
The first generation of the Gemini AI having been unveiled in December 2023, expectations are logically focused on an announcement during this month of December, rightly or wrongly.
And as if to support the idea of an imminent launch, clues suggest that Gemini 2.0 would already be more or less accessible to certain users, at least in an experimental version.
A new model Gemini-2.0-Pro-Exp-0111 was in fact spotted in a subsection of Gemini Advanced, the most powerful version of Google’s artificial intelligence, with the subtitle “our new experimental model“.
Gemini 2.0 for against OpenAI’s o1 model
It is difficult to know whether this is Gemini 2.0 or a test version of a current AI model and whether it is reserved for internal experimentation or accessible more widely.
A user indicates that this new model would perform better on a test than the o1-mini model of OpenAI, one of the advanced versions of GPT-4 simulating human-like reasoning ability and intended to pave the way for AGI, artificial general intelligence.
An insider, Jimmy Apples, also suggests that Gemini 2.0 would already be deployed to certain professional customers from Google, potentially to get initial feedback before a global launch.
Gemini 2.0 will be fully multimodal and should take advantage of the latest advances to interpret written, voice and image/video queries thanks to enhanced reasoning capabilities in planning and problem solving, and expanded token ranges.
Morocco
