Google introduces the Gemini Flash 8B, a compact closed model with the same core capabilities as full models, such as multimodal support for both audio and visual inputs, supporting up to 1 million tokens. The Flash-8B was actually revealed in Gemini 1.5 reports since mid-last year, indicating it is still in development. Test scores show a decrease compared to Gemini 1.5 Flash, falling short when compared to Llama 3.1 8B in most tests. However, features like multimodal and context window provide diverse usability options. Converting Thai speech to text with Gemini Flash-8B yielded precise results.
In addition to Gemini Flash-8B, Google has released experimental versions of Gemini 1.5 in both Pro and Flash, labeled as version 0827. Despite the models’ small size for easy organizational use, Google distinguishes between models for commercial purposes under the Gemini name and the clearly open-source Gemmma models. Currently, Gemini Flash-8B is available for free use in Google AI Studio.
TLDR: Google unveils the Gemini Flash 8B, a compact model with advanced capabilities. Gemini models are separated for commercial use under the Gemini name from open-source Gemma models. Gemini Flash-8B is now free to use in Google AI Studio.
Leave a Comment