Google has just launched Gemini 2.5 Flash, a hybrid reasoning AI model in preview that significantly improves over its predecessor and rivals. Gemini 2.5 Flash offers reasoning boosts, a controllable 'thinking budget' (up to 24k tokens), and strong performance in reasoning, STEM, and visual benchmarks — all at a lower cost compared to competitors.
The model is now available via API through Google AI Studio and Vertex AI, and is also integrated experimentally within the Gemini app. Users can toggle the reasoning depth based on use cases, enabling fine-grained control over cost, quality, and speed — a step toward affordable large-scale applications that demand smarter response tuning.