Google published a roundup of its most notable artificial intelligence announcements during August 2026, covering new models, consumer devices, and tools aimed at students, developers, and researchers. This is part of the company’s effort to move Gemini from being a conversational assistant to an integrated layer across devices, search, and work tools.
Lower-Cost Models and More Accurate Voice Tasks
Google announced the launch of Gemini 3.7 Flash as a model aimed at programming, building software agents, knowledge workflows, and web development. According to the company, the model arrived just three weeks after Gemini 3.6 Flash, with an introductory price equal to half the previous model’s cost per million tokens. This point is important for developers balancing model quality against the operating cost of repetitive tasks, but the material does not provide independent performance comparison figures or complete details about how long the introductory price will last.
Google also unveiled Gemini 3.5 Transcribe, a real-time speech-to-text model targeting voice agents, live captions, and post-call analytics. The company says the model handles noise and specialized terminology by converting raw audio into text that is more accurate and context-aware.
Gemini Moves Into Devices and Work Tools
At the Made by Google 2026 event, the company unveiled the Pixel 11, Pixel 11 Pro, Pixel 11 Pro XL, and Pixel 11 Pro Fold. The devices include the Google Tensor G6 chip, which runs the latest Gemini Nano model, along with improvements to the cameras and durability. According to Google, the series was designed to deliver what the company calls “Gemini Intelligence,” meaning personal assistance integrated into the phone experience.
In Gemini Live, Google added tools such as Personal Intelligence, Daily Brief, Spark, and hands-free email inbox management. In practice, these additions reflect the app’s shift from answering questions to carrying out or organizing daily tasks, while feature availability remains dependent on countries, languages, and subscriptions.
Education and Media Generation
Google said it is offering a free Google AI plan for one year to eligible college students worldwide, along with a student hub, tools for educators, and an SAT preparation feature. It added AI-based educational features to search, including interactive visual explanations, the creation of practice tests for the SAT, ACT, GRE, and LSAT, and step-by-step learning through Lens, as well as content organization in notebooks.
The company also announced Gemini Omni 1.1 Flash for video generation with greater control, including scene extension, interpolation between the first and last frames, 4K upscaling, and faster prototyping. The model is available in Google Flow, Google AI Studio, Gemini Enterprise Agent Platform, and the Gemini app.
From One Billion Users to Scientific Applications
According to Google, the Gemini app has surpassed one billion monthly users, with 63% of users speaking to it directly, while the app generates more than 150 million images daily. The company also says small businesses use image, video, and audio creation tools to produce marketing materials.
Beyond consumer products, Google noted that downloads of the open-source Gemma models have surpassed one billion, with uses extending from phones and edge infrastructure to space. It also presented Operation Blue Skies, a collaboration with the United Kingdom government and the aviation sector to use forecasts to adjust routes and avoid the formation of contrails. In the climate field, the company said a paper in Nature demonstrated WeatherNext 2’s ability to predict hurricane tracks, intensity, and wind structure, and announced that it was making the model available as open source to researchers.
Why Does This News Matter?
This roundup shows that the competition is not centered on a single model, but on reducing usage costs, connecting artificial intelligence with devices and services, and expanding its presence in education, media, and science. However, most of the figures and results cited come from Google itself, and feature availability varies by country, language, and age; some functions may require a subscription or higher usage limits. Independent comparisons and real-world performance tests therefore require separate review.