November 19th, Beijing time is early this morningGoogleFormally new Gemini 3 Series model and position it as the "most intelligent" "fact-based" AI system to date. This version is also seen as a key opportunity for Google to regain its lead position in the consumer-oriented AI competition, along with the down-line GPT-5。
All Google users can directly use the new flagship model Gemini 3 Pro in the Gemini application and introduce it in the search subscription service simultaneously. Google DeepMind Senior Director Tulsee Doshi stated that, as the search pattern evolved, Gemini 3 Pro would bring Google closer to the goal of making information accessible and really useful。
Doshi said, "The real important step is to move away from the old model, which can only give a text response, and to offer oneMore complete, richer presentation. "
Gemini 3 Pro's core capability is "Natural Multiple Modes"One-time processing of text, pictures and audioinstead of split into different processes. Google, for example, can modelMake a recipeIt is also possible to automatically generate interactive cards based on multiple course videos。
The upgrades brought by the model have permeated a range of Google products. In the Canvas workspace of the Gemini application, users can build more fully functional programs using new models. The " Generating Interface" function tested in Gemini Labs allows Gemini 3 Pro to generate a magazine-style visual layout based on a hint, or to create a dynamic interface that adapts to demand。
Gemini 3 Pro will also present the results in a more intuitive manner, including pictures, tables, grids and even simulations, in the AI Mode search. Based on the upgraded "Query Fan Out" technique, the model not only disassembles the problem into searchable details, but also better understands the intent to ask questions, and then finds the elements that may have been omitted in the past。
According to the media, The Verge, Google also sarcasmed OpenAI, highlighting Gemini 3 Pro Not like ChatGPTI don't know. Doshi says that the answer to the new model will be “changed and visible”, and Google says Gemini 3 Pro's expression is “better, more accurate, more direct, and a real insight replaces a guest and a suit - telling the user what to know, rather than responding to what the user wants to hear”. Google says that the new edition's performance in the area of “sweetness” has been significantly reduced, and ChatGPT had to correct similar problems earlier this year。
According to 1AI, Gemini 3 Pro also has significantly enhanced its reasoning and intelligence capabilities to respond to complex tasks and haveStable long-term planning capacityI don't know. The Gemini smart body function, which is being tested in the Gemini application, is being driven by this model, which can be used as a substitute for the user to organize mail, or to search for and schedule a trip。
Currently Gemini 3 Pro is on a widely used assessment platform LMArena Top. Its Deep Think model further enhances its reasoning capability, but is now open only to security testers。
Gemini 3 Pro has been made available to all users in the Gemini application, and Google AI Pro and Ultra subscribers in the U.S. can also choose "Thinking" in the search for AI Mode. Gemini smart body will be first available to AI Ultra subscriber。