Google just officially introduced the stable versions of their latest AI models Gemini 2.5 Pro and Gemini 2.5 Flash. Additionally, they unveiled a new, lightweight model called Flash Lite.
After months of testing, the flagship AI model, Gemini 2.5 Pro, was released to the public. It initially entered the testing phase on March 25, 2025, only available to paying users and developers. However, just four days later, Google surprisingly made the model accessible to free users as well.
Prior to the I/O conference, the same model received significant enhancements in coding, with the latest update coming at the beginning of June. The recent announcement of the stable version did not include any changes from the June preview version, but there were adjustments made to limits and prices.
In addition, Google introduced a new model called Gemini 2.5 Flash Lite.
Google continues to maintain a tiered approach to their AI models. Free Gemini app users will have limited access to 2.5 Pro, while AI Pro subscribers will have extended access with a limit of 100 prompts per day. The highest level of access is offered through the Google AI Ultra package. Model 2.5 Pro is optimized for tasks requiring „reasoning, mathematics, and coding“, while 2.5 Flash is designed for „fast, versatile assistance“.
Gemini 2.5 Flash is now available to developers with updated pricing: $0.30 for 1 million input tokens (an increase from the previous $0.15) and $2.50 for 1 million output tokens (a decrease from the previous $3.50). Google removed the pricing difference between „thinking“ and „non-thinking“ modes, maintaining a uniform pricing level regardless of the size of input tokens.
Google also introduced a preview version of Gemini 2.5 Flash Lite, designed for high-volume tasks sensitive to latency, such as translations and classifications. This model offers lower latency than 2.0 Flash-Lite and 2.0 Flash at the same cost as its predecessors.
Flash Lite supports multimodal input, a context window with 1 million tokens, and native tools including Google Search, code execution, and URL context. The model exhibits higher quality than 2.0 Flash-Lite in all areas including coding, mathematics, science, reasoning, and multimodal benchmarks. It is available through Google AI Studio, API, Vertex AI, and is expected to be part of the Gemini app.