Google accelerates AI race: new Gemini 3 Flash model enters search and business tools
18 December 21:08
Google has introduced a new artificial intelligence model, Gemini 3 Flash, the fastest in its lineup. It is focused on everyday tasks, as well as working with photos and videos, and has already become the basic model in the Gemini app and is built into Google search. This was announced by Google, "Komersant Ukrainian" reports.
Why it is important
The launch of the Flash version demonstrates Google’s desire to strengthen its position in the competitive AI race by relying not only on “power” but also on speed, cheapness, and scalability. Low cost of use and high performance make the model attractive to businesses and developers.
What Gemini 3 Flash can do
The model allows you to upload images, thumbnails, or videos and receive detailed explanations or recommendations. One of the key innovations is the ability to generate mobile app prototypes based on a text description only.
According to Google, the Flash version consumes on average 30% fewer tokens than Gemini 2.5 Pro, while maintaining focus on complex tasks longer.
Figures and tests
Gemini 3 Flash scored 90.4% in the GPQA Diamond test and 33.7% in the Humanity’s Last Exam, which are scores that allow it to compete with larger and more expensive models. The cost of use is about $0.5 per 1 million input tokens.
Where it has already been integrated
The model already works in Figma, Cursor, Latitude, and JetBrains, which demonstrates its focus on developers and corporate processes. Google is also expanding the ecosystem around AI, including through the Opal platform for creating apps without code and NotebookLM for working with documents.
The release of Gemini 3 Flash underscores the shift in focus in the AI industry: from demonstrating “maximum capabilities” to practical, fast, and cost-effective solutions. Amid discussions about the profitability of large AI projects, such as OpenAI, Google signals that it is betting on efficiency and mass adoption.
