Introducing Gemini 2.5 Flash

To today’s fascinating technical insight, we have a brand new technical blog post that dives into some exciting and fascinating topics. The title of our tech blog post is “Today We Are Rolling Out an Early Version of Gemini 2.5 Flash in Preview Through the Gemini API Via Google AI Studio and Vertex AI.” Our goal with this new version of Gemini 2.0 Flash is to introduce a more thinking-capable model that prioritizes reasoning capabilities while still prioritizing speed, cost, and latency. To achieve this goal, we’ve also introduced a new thinking model called Gemini 2.5 Flash with thinking capabilities that allows developers to perform faster, more accurate thinking before responding. This thinking process enables the model to better understand complex tasks, analyze research questions, and generate more comprehensive answers. Gemini 2.5 Flash is now available in preview through Google AI Studio and Vertex AI. To give developers flexibility, we have enabled setting a thinking budget that offers fine-grained control over the maximum number of tokens a model can generate while thinking. This means that developers can choose the amount of reasoning required for a given prompt, allowing them to improve quality, cost, and latency in different use cases. We’ve also enabled setting a specific token budget for the thinking phase using a parameter in API calls or in the slider in Google AI Studio. We also have fine-grained control over how much we want our model to think, which allows us to set a higher budget for tasks that require more thorough reasoning while allowing less time and resources to be used on tasks that are easier to complete. The budget can range from 0 to 24576 tokens in our API calls or between 0 and 24576 tokens in Google AI Studio, allowing us to choose the amount of time we want to spend thinking for a given prompt. In addition to our new model capabilities, our developers can also set aside more resources to be used on tasks that require more thorough reasoning in a higher budget while allowing less time and resources to be spent on tasks that are easier to complete. We’ve also introduced some code examples and thought guide for those who want to explore how thinking can help solve more complex problems. Before we move onto our next topic, we would like to emphasize the importance of maintaining quality and continuously improving performance in all aspects of development. To achieve this goal, we’ve introduced a new API reference and thinking guide in our developer docs that offer detailed information on how controllable reasoning can help developers solve more complex problems. In conclusion, we encourage you to get started with our newly released model capabilities and explore how they can help you solve more complex problems. If you have any questions or need further assistance, please don’t hesitate to contact us. Thank you for your interest in Gemini 2.5 Flash and our developer resources.