Page Loader
Summarize
Google rolls out speedy Gemini 1.5 Flash AI model
Gemini 1.5 Flash boasts the same multimodal capabilities as the Pro version

Google rolls out speedy Gemini 1.5 Flash AI model

May 14, 2024
11:31 pm

What's the story

Google has unveiled its latest addition to the Gemini family of AI models, the Gemini 1.5 Flash, at its annual I/O conference for developers. The new model is designed for speed and efficiency, excelling in tasks such as chat applications, image and video captioning, summarization, and data extraction from long documents and tables. This info was shared by Demis Hassabis, CEO of Google DeepMind.

Optimization

Gemini 1.5 Flash: A lighter, cost-effective AI model

The development of Gemini 1.5 Flash was motivated by developers' demand for a lighter and less costly model than the Pro version launched in February. Despite its lightweight nature, Gemini 1.5 Flash retains the same power as its predecessor. This was accomplished through "distillation," a procedure where essential knowledge and skills from Gemini 1.5 Pro, were transferred to the smaller model, as explained by Hassabis.

Model performance

Gemini 1.5 Flash's impressive multimodal capabilities

Gemini 1.5 Flash boasts the same multimodal capabilities as the Pro version, capable of processing one million tokens simultaneously. This means it could analyze a 1,500-page document or a codebase with over 30,000 lines at once. While not designed for consumers, these models give developers a faster and cheaper way to build their own AI products and services using Google's technology.

Upgrades

Google enhances Gemini 1.5 Pro

In addition to launching Gemini 1.5 Flash, Google is also enhancing the abilities of Gemini 1.5 Pro to write code, reason, and also parse audio and images. The tech giant plans to double the Pro model's current context window to two million tokens later this year. This upgrade will allow it to process two hours of video, 22 hours of audio, over 60,000 lines of code, or over 1.4 million words simultaneously.