Google has introduced a significant enhancement to its flagship AI models with the addition of thought summaries for both Gemini 2.5 Pro and Flash models, available through the Gemini API and Vertex AI.
Thought summaries transform the model's raw thinking process into a clear, structured format with headers, key details, and information about model actions, such as when they use tools. This feature represents a major step forward in making AI reasoning more transparent and debuggable.
"We hope that with a more structured, streamlined format on the model's thinking process, developers and users will find the interactions with Gemini models easier to understand and debug," Google stated in its announcement at Google I/O 2025.
The thought summaries capability comes alongside other enhancements to the Gemini 2.5 family, including the extension of thinking budgets to Gemini 2.5 Pro. This feature, previously available only in Flash, allows developers to control the number of tokens a model uses to think before responding, or even turn thinking capabilities off entirely, providing greater control over cost, quality, and latency tradeoffs.
For enterprise users, thought summaries represent a breakthrough for trust and steerability. Organizations evaluating AI output for critical tasks—like reviewing policy implications, coding logic, or summarizing complex research—can now see how the model arrived at its conclusions, making AI decision-making more transparent to stakeholders.
Google has announced that Gemini 2.5 Flash is now available in preview in Google AI Studio for developers, in Vertex AI for enterprises, and in the Gemini app for all users. The updated version will be generally available for production in early June, with Gemini 2.5 Pro following soon after.