Today, Google has officially launched two new models, Gemini-1.5-Pro-002 and Gemini-1.5-Flash-002. Within Google's series of models, Gemini Pro is categorized as a medium-sized model available for paid users. In contrast, Gemini Flash is distilled from Gemini Pro and first debuted at Google I/O in May this year; it is now available for free use within Gemini, with developers also receiving a limited free API usage quota.
The names Gemini-1.5-Pro-002 and Gemini-1.5-Flash-002 indicate that this update is not a major version release but rather an overall enhancement of the existing model framework.Notably, the input token price for Gemini 1.5 Pro has been reduced by 64%, while the output token price has decreased by 52%. The cost for incremental cached tokens has also been lowered by 64% for prompts under 128K tokens, effective from October 1, 2024. This, combined with context caching, will further decrease the costs associated with building applications using Gemini.
Additionally, the rate limit for 1.5 Flash has been increased from 1,000 RPM to 2,000 RPM, while the rate limit for 1.5 Pro has been raised from 360 RPM to 1,000 RPM. These changes will come into effect over the next few weeks. On the performance front, Google has reported that the 1.5 Flash model now boasts double the output speed and three times lower latency compared to previous versions.
Importantly, Google has stated that the newly released Gemini models will not have filters applied by default, allowing developers to configure the settings according to their specific use cases. However, Gemini will continue to provide a range of safety filters that developers can apply as needed to their applications.
Furthermore, Google has indicated that the latest models have made notable advancements in areas such as mathematics, handling long-context windows, and visual processing capabilities.
This release signifies Google’s ongoing commitment to enhancing the functionality and flexibility of its AI models, providing developers with more robust tools to create innovative applications.