Gemini 1.5 Flash has been released in a smaller and faster version. Google is making Gemini 1.5 Flash-8B available for a smaller price as well.
From the announcement of Gemini 1.5 Flash-8B, this smaller model seems to be a better option than the original Gemini 1.5 Flash. The new model comes at half the price and doubles the speed at which output is generated.
Size is not a priority on the smartphone
Typically, smaller models are less interesting because performance decreases, which means only simple tasks remain feasible. In this case, this is not a disadvantage. Gemini 1.5 Flash was presented in the spring as the smallest version within the Gemini 1.5 family. It has a clear use case: the smartphone market. AI workloads for this scenario are, by definition, small due to the hardware only being suitable for smaller performance compared to a workstation, for example.
For these less complex tasks, Gemini 1.5 Flash-8B appears to be more responsive than Gemini 1.5 Flash, writes Gemini API Senior Product Manager Logan Kilpatrick in the announcement. Benchmark testing shows the model is well suited for tasks such as chatting, converting speech to text, and translating long texts.
Available to developers
The announcement is aimed at developers. They can access the model through the Google AI Studio and the Gemini API. The model can process up to 4,000 requests per minute. Use is free until Oct. 14. The price scheme is already available and, according to Kilpatrick, the lowest Google charges within the Gemini family. An overview can be found below.
Tip: Gemini Live will soon offer recipes in 40 different languages