On June 17, 2025, Google officially launched its Gemini 2.5 AI models, moving the 2.5 Pro and 2.5 Flash to stable versions and introducing the preview of 2.5 Flash-Lite. The Gemini family is now available in platforms like Google AI Studio and Vertex AI, featuring a one-million-token context and improved developer controls. Flash-Lite is optimized for high-volume, low-latency tasks, delivering answers in under 100 milliseconds while reducing costs per token. It exceeds its predecessor in coding and math benchmarks, requiring 20-30% fewer tokens. The 2.5 Pro model debuts Deep Think, a feature that evaluates multiple hypotheses for complex tasks, alongside enhancements like native audio output and increased security against prompt injections. This update reflects a strategic balance of speed and capability, with companies like Spline and Snap already utilizing the technology in production.
Source link
