Google launches its new gemini 2.0 models: here is all that changes

WhileSiri is late, Google gets ahead by developing its family of AI models,Gemini. A renewal that materializes at high speed on user screens. Initially deployed on the web platform last Thursday,The new versions quickly reached mobile applicationsAndroid and iOS. An unusually fast calendar for Google, which usually prefers to spread its deployments over several weeks. As of Saturday afternoon, many accounts already benefited from improvements.

From now on,A palette of five different models are available. Gemini 2.0 Flash for daily use and have quick responses, 2.0 Flash Thinking for more complex and in -depth reasoning, Deep Research, a model that can analyze colossal amounts of data, perzonization that takes into account the history of the user to refine its responses, and 2.0 Pro, which excels in various fields.

Faster models accessible to all

The most striking change concerns generalized access to the Flash Thinking model. Previously reserved for premium subscribers, this model becomes accessible to users of the free version.It is much more efficient and fasterIn the execution of the tasks that the flash model and even displays its reasoning when asked a question. Google hasdiscreetly modified his nomenclature: the term«experimental»Now appears in tiny and parentheses, both for Flash Thinking and for the Pro model.

The overall experience also gains in fluidity and Gemini no longer works in a partitioned manner. The firm has simplified access to other applications of its ecosystem: the specific model "with apps», Introduced last month to connect Gemini to Gmail or YouTube, disappearsfor the benefit of direct access via Flash Thinking. Navigation is, therefore, much more pleasant and fluid.

At the same time, importing documents becomes accessible to all users, including those of the free version. SubscribersGemini Advancedbenefit from a considerable improvement in contextual memory, which now reaches a million tokens. This means that the model can remember a much larger amount of information exchanged during a conversation,which allows much longer, nuanced and detailed dialogues.

Gemini had the dirty tendency to have short memory, unlike a certainChatGPT; The expansion of his memoryshould do him the greatest good.

One of the most impressive models is certainly Deep Research, capable of paying complex research and searching the web at full speed and giving its sources on a subjectx. He can then provide a detailed synthesis, point by point, while justifying his reasoning. It is a little longer to relaxation than the others, but the information he later distills is much more in -depth.

Example below, with the question: "Can you make me the history of Apple's creation and its evolution until today?». Gemini will first offer you aresearch plan,that you can modify and validate so that it begins to seek.

The model shows you its reasoning, the sources it uses to generate its syntheses andThen produces an extremely complete and well -organized summary(see below). On the few essays we have done, absolutely no error was noted despite choices of sharp subjects (physical geography, zoology, history, genetics, etc.).

The model called Personalization (unavailable in web/mobile version when writing this article)draw from your research history to contextualize your answers. This model, still at the experimental stage, could find its relevance to use for travel recommendations, multimedia content ideas to visualize or suggestions for new skills to acquire.

Google obviously tries to stand out with a new strategy:the diversification of its modelsby repositioning them halfway between accessibility and performance. A big advantage for those who want to find a shoe in their foot by choosingThe most appropriate modelDepending on its requirements or according to the task to be performed. We hope that Apple will take seed forApple Intelligence.

• Google accelerates the deployment of its new Gemini models, now offering five variants adapted to different uses, with extended accessibility to free users.
• The user experience improves with more fluid navigation, the deletion of barriers between applications, extended contextual memory and the integration of advanced functionalities such as imports of documents.
• With Deep Research for in -depth and personalization analyzes for responses adapted to research habits, Google is betting on a more efficient and customizable AI.