Meet the Newest Member of Google’s Gemini 2.5 AI Family
Google is adding a new lightweight variant to its growing Gemini 2.5 family of AI models: Gemini 2.5 Flash-Lite . The model is now available in preview, giving developers early access to test its capabilities. At the same time, Google has announced that the more powerful Gemini 2.5 Pro and Gemini 2.5 Flash are now generally available for production use.
According to Google, Flash-Lite outperforms earlier Gemini 2.0 versions in areas like coding, math, and logical reasoning. It’s specifically designed for high-volume tasks where speed matters most — such as language translation, content classification, or data filtering at scale.
To match this use case, pricing for Flash-Lite is significantly lower than other models in the lineup. At just $0.10 per million input tokens and $0.40 per million output tokens , it’s by far the most cost-effective option. However, when it comes to handling complex reasoning challenges, its baseline performance remains limited. On a challenging benchmark known as Humanity’s Last Exam , Flash-Lite scored only 5.1% , indicating room for improvement in advanced cognitive tasks.
⚡ The “Thinking” Mode: A Game-Changer for Performance
One of the standout features of Gemini 2.5 Flash-Lite is its new “Thinking” mode . When activated, the model takes extra time to process prompts, which leads to a noticeable boost in performance.
For example:
- On the AIME 2025 math benchmark , accuracy jumps from 49.8% to 63.1%
- In document search and recall tests, the model nearly doubles its effectiveness
This mode allows developers to trade off speed for smarter responses, making it ideal for applications where accuracy matters more than immediacy.
🛠️ Availability and Use Cases
Developers can start experimenting with Gemini 2.5 Flash-Lite today through Google AI Studio and the Vertex AI platform . Meanwhile, the full versions of Gemini 2.5 Flash and Pro are already available on those platforms, and have also been integrated into the main Gemini app for consumer use.
Interestingly, Google revealed that customized versions of both Flash and Flash-Lite are already being used internally to enhance parts of Google Search , suggesting these models are playing a growing role in the company’s core services.
