Google just rolled out an updated version of Gemini 2.5 Flash at the biggest conference of the year, I/O 2025, on May 20.
Google DeepMind head Demis Hassabis stated that the recently launched and improved 2.5 Flash is “better in nearly every dimension.”
Gemini 2.5 Flash is said to be a 22% more efficient and secure model.
With an efficient variant, users will now receive the similar responses but using lesser tokens. In addition, the Alphabet-owned company is adding "Thinking budgets," allowing you to set the number of tokens it can use prior to responding.
Availability
Gemini 2.5 Flash will be accessible to the public in June.
According to Demis, It is one of the most intelligent AI models all across the globe. It can simulate entire cities, with a range of advanced capabilities for learning and more.
Text-to-speech feature
Gemini 2.5 has received the latest text-to-speech feature in a more natural tone, and it can even do things like whisper. Interestingly, the model can switch between different languages using the same voice.
