Google aims for the throne of artificial intelligence with Jumnai 3 Pro

With the official introduction of Jumnai 3, Google took a big step towards taking the title of artificial intelligence leader. With the unveiling of its new collection of artificial intelligence models, which includes the flagship model “Jumnai 3 Pro”, this company claims to have built the smartest and most accurate artificial intelligence systems in its history. Google’s goal with this release is to directly compete and even surpass OpenAI’s GPT-5 model in the consumer market. The good news for users is that access to Jumnai 3 Pro is available for everyone from today in the exclusive Jumnai application.

DeepMind CEO Tulsi Doshi emphasized at the event that Jumnai 3 Pro brings Google closer to its ultimate goal of “organizing the world’s information and making it accessible and useful for everyone.” He considered one of the outstanding features of this model to be leaving the framework of simple text answers and providing a richer and multifaceted experience. Jumnai 3 is inherently a multimodal model and is able to process text, sound and image at the same time. This feature allows users to build more complete apps in tools like Canvas, or use the Generative Interfaces feature to create dynamic, magazine-like visual templates.

Jumnai 3 Pro has also set new records in the field of technical performance. This model scored 1501 points in the prestigious LMArena benchmark, which is a significant improvement over the 1451 points of the previous generation. It also scored 37.5% and 91.9% unaided on complex reasoning tests such as Humanity’s Last Exam and GPQA Diamond. In a taunt to the competition, Google claims that Jumnai 3’s responses are “intelligent and unflattering” and deliver exactly what the user needs. In addition, a new experimental feature called Gemini Agent has been introduced, which can perform various tasks in the application on behalf of the user.

Source: 9To5Google