Google has expanded its family of artificial intelligence models by launching three new versions of the Gemini 2.0 model: the basic version Gemini 2.0 Flash, the economical version Gemini 2.0 Flash-Lite, and the experimental version Gemini 2.0 Pro. These new models are designed to meet various usage needs while providing a diverse balance between performance and cost.
The basic version Gemini 2.0 Flash was first launched last December and is now fully available, featuring higher usage frequency limits and improved performance. Meanwhile, Gemini 2.0 Flash-Lite is a low-cost variant aimed at developers, currently in public preview via API.
Gemini 2.0 Pro is an experimental model designed for complex prompts and coding tasks, with a context window extended to 2 million tokens, which is double that of the Flash version.
Currently, these models only support text output, but Google plans to add image, audio, and live video capabilities to the Flash and Pro models in the coming months. Additionally, all three models can handle images and audio as input.
In testing, Google benchmarked the performance of Gemini 2.0 Pro, showing that it outperformed previous models in almost all areas. In mathematical tasks, Gemini 2.0 Pro scored 91.8% on the MATH benchmark, while scoring 65.2% on HiddenMath, far exceeding the performance of the Flash version.
Gemini 2.0 Flash scored 29.9% in OpenAI's SimpleQA test, while the Pro model scored 44.3%. This indicates that Gemini 2.0 Pro has higher accuracy in answering complex questions.
It is worth noting that Google has adjusted its API pricing by removing the distinction between short and long text queries. This means that the costs for mixed workloads (text and images) may be lower than those for the Gemini 1.5 Flash version, despite improved performance. Specifically, the pricing for Gemini 2.0 Flash is set at $0.075 per million inputs and $0.40 per output. The relatively cheaper Gemini 2.0 Flash-Lite charges $0.075 for text, image, and video inputs, and $0.30 for text outputs.
Although the price of Gemini 2.0 Flash is higher than that of its predecessors, the new Flash-Lite model offers better performance at the same price, aimed at bridging the gap between price and performance.
Additionally, Google has updated the Gemini application, fully opening the features of the Gemini series to provide users with a richer experience.
All models are available through Google AI Studio and Vertex AI, as well as Google's advanced Gemini Advanced chatbot on both desktop and mobile devices.
Key Points:
🔹 Google has launched three new Gemini 2.0 models, including Flash, Flash-Lite, and Pro, catering to different needs.
🔹 Gemini 2.0 Pro excels in math and accuracy tests, scoring significantly higher than previous models.
🔹 API pricing adjustments make the cost of mixed workloads more competitive, while Flash-Lite fills the market gap with better performance.
