Google releases Gemini 2.5 Flash with hybrid reasoning and boasts faster speeds
What you need to know
- With Hybring Reasoning, developers can enable or disable the ‘thinking’ process, maintaining speed while improving accuracy.
- The new Gemini model claims to be faster and more cost-effective than previous versions.
- Gemini 2.5 Flash also outperforms competing AI models in benchmark tests across various tasks.
Late last week, Google shared that it is releasing the Gemini 2.5 Flash — a new version of the Gemini AI model. It highlights the hybrid reasoning capability, which further allows developers to control the level of reasoning applied to a given prompt.
In the announcement post, Google notes that developers can now start building with the new Gemini 2.5 Flash model, which is also a faster and cheaper model. It is released in preview through the Gemini API via Google AI Studio and Vertex AI.
In addition to being cost-effective, the new model is also a low-latency thinking model that is reasonably priced, particularly considering its solid benchmarks. This is because the new model is based on the popular foundation of Gemini 2.0 Flash.
As mentioned, it is a hybrid model, which allows developers to turn on thinking or disable it. Despite disabling it, the company promises that it can still maintain the fast speeds of 2.0 Flash and simultaneously improve performance.
As for the thinking model, the Gemini 2.5 models are capable of reasoning before responding. This means it doesn’t immediately generate an output; “the model can perform a ‘thinking’ process to better understand the prompt, break down complex tasks, and plan a response.” The complex tasks here can include solving math problems or analyzing research questions. The thinking process claims to allow the model to give more accurate and comprehensive answers.
Google has also shared the Gemini 2.5 Flash model’s benchmark tests in comparison with some of the competition, which includes tasks based on mathematics, science, code generation, code editing, and visual reasoning — and does outperform the rival AI models.
Other than the developers taking advantage of the new Gemini 2.5 Flash, regular consumers can also try it out in their Gemini app, alongside Gemini 2.0 Flash Thinking and Gemini 2.0 Pro Experimental models, which were unveiled in February this year.
Post Comment