At the I/O 2026 conference, the tech giant unveiled Gemini 3.5 Flash—a new model for the AI agent era that was made available to all users for free and proved to be four times faster than its competitors.
California, USA. At the annual Google I/O 2026 developer conference, Alphabet CEO Sundar Pichai announced the launch of Gemini 3.5 Flash—the first model in the updated Gemini 3.5 lineup. The new model is positioned as the foundation for the so-called "era of agentic AI," offering a previously unattainable balance between deep intelligence and task execution speed.
The biggest surprise of the presentation was that the lightweight Flash version outperformed the previous flagship model, Gemini 3.1 Pro, in key industry benchmarks. Specifically, the model demonstrated outstanding results in complex code generation, tool use, and multi-step logical reasoning. At the same time, Gemini 3.5 Flash’s token generation speed proved to be four times faster than that of other comparable advanced AI systems on the market.
"If large companies move even 80% of their current AI workloads to Gemini 3.5 Flash instead of heavier systems, they could save over $1 billion a year. That is a massive cost optimization," Sundar Pichai emphasized during his keynote presentation.
Key characteristics and technical capabilities of Gemini 3.5 Flash:
Built-in multimodality: the model effectively processes not only text and images but also audio and video files right out of the box, delivering high-quality responses.
Optimization for AI agents: Gemini 3.5 Flash is tailored for executing long-horizon tasks, iterative programming, and deploying sub-agents.
Immediate availability: the model has officially reached General Availability (GA) status and has already been implemented as the primary (default) model in the Gemini consumer app, as well as in the AI Mode of Google Search.
A pleasant surprise for general users is that the capabilities of Gemini 3.5 Flash can be tried out completely free of charge via the Gemini web version and mobile apps (subject to daily generation limits). For developers and companies, paid API access is available through Google AI Studio and Vertex AI.
As for the more powerful version—Gemini 3.5 Pro—Google has confirmed that it is already undergoing active internal testing and will be made available to the general public next month (June 2026).
0 Comments