Gemini 2.5 Update: Flash-Lite Writes OS in Real Time
Gemini 2.5 update: Flash-Lite debuts as cost-efficient AI, Pro excels in coding. New pricing & features.
"AI Disruption" Publication 6900 Subscriptions 20% Discount Offer Link.
The Gemini series models just received a wave of updates:
Gemini 2.5 Pro stable version is released and fully available, unchanged from the June 5 preview.
Gemini 2.5 Flash stable version is released and fully available, unchanged from the May 20 preview, but with updated pricing.
Gemini 2.5 Flash-Lite has been newly launched and is available in preview.
Google CEO Sundar Pichai tweeted that the newly introduced Gemini 2.5 Flash-Lite is the most cost-effective model in the 2.5 series.
Google positions 2.5 Flash-Lite for “high-volume, cost-efficient tasks.” In contrast, 2.5 Pro is suited for programming and high-complexity tasks, while 2.5 Flash strikes a balance, ideal for daily tasks requiring faster speeds.
According to the official blog, Gemini 2.5 Flash-Lite is an inference model supporting multimodal inputs, a 1 million token context window, and dynamic control of thinking budget via API parameters. Optimized for cost and speed, its “thinking” function is disabled by default, unlike other Google DeepMind models.
Beyond function calling, 2.5 Flash-Lite supports all existing native tools, such as Google Search-based grounding, code execution, and URL context.
Performance-wise, 2.5 Flash-Lite is slightly weaker than 2.5 Flash, with the thinking-enabled version outperforming the disabled one. However, it surprisingly shows slight score advantages in specific metrics like AIME 2025 and FACTS Grounding.