Google’s new Gemini 1.5 Flash AI mannequin is lighter than Gemini Professional and extra accessible

Google introduced updates to its Gemini family of AI fashions at I/O, the corporate’s annual convention for builders, on Tuesday. It’s rolling out a brand new mannequin known as Gemini 1.5 Flash, which it says is optimized for velocity and effectivity.

“[Gemini] 1.5 Flash excels at summarization, chat purposes, picture and video captioning, information extraction from lengthy paperwork and tables, and extra,” wrote Demis Hassabis, CEO of Google DeepMind, in a blog post. Hassabis added that Google created Gemini 1.5 Flash as a result of builders wanted a mannequin that was lighter and cheaper than the Professional model, which Google announced in February. Gemini 1.5 Professional is extra environment friendly and highly effective than the corporate’s unique Gemini mannequin introduced late final yr.

Gemini 1.5 Flash sits between Gemini 1.5 Professional and Gemini 1.5 Nano, Google’s smallest mannequin that runs regionally on gadgets. Regardless of being lighter weight then Gemini Professional, nevertheless, it’s simply as highly effective. Google stated that this was achieved via a course of known as “distillation”, the place probably the most important data and abilities from Gemini 1.5 Professional have been transferred to the smaller mannequin. Which means Gemini 1.5 Flash will get the identical multimodal capabilities of Professional, in addition to its lengthy context window – the quantity of information that an AI mannequin can ingest directly – of 1 million tokens. This, based on Google, signifies that Gemini 1.5 Flash shall be able to analyzing a 1,500-page doc or a codebase with greater than 30,000 traces directly.

Gemini 1.5 Flash (or any of those fashions) aren’t actually meant for shoppers. As an alternative, it’s a quicker and cheaper means for builders constructing their very own AI services and products utilizing tech designed by Google.

Along with launching Gemini 1.5 Flash, Google can be upgrading Gemini 1.5 Professional. The corporate stated that it had “enhanced” the mannequin’s talents to put in writing code, motive and parse audio and pictures. However the largest replace is but to come back – Google introduced it is going to double the mannequin’s current context window to 2 million tokens later this yr. That will make it able to processing two hours of video, 22 hours of audio, greater than 60,000 traces of code or greater than 1.4 million phrases on the identical time.

Each Gemini 1.5 Flash and Professional at the moment are out there in public preview in Google’s AI Studio and Vertex AI. The corporate additionally introduced right now a brand new model of its Gemma open mannequin, known as Gemma 2. However except you’re a developer or somebody who likes to tinker round with constructing AI apps and providers, these updates aren’t actually meant for the typical shopper.

Make amends for all of the information from Google I/O 2024 proper here!