Google announces Gemini Flash 1.5 lightweight AI model at Google I/O

Google is using its annual developer conference to showcase what the company calls its lightest and most efficient AI models.

At Google I/O on Tuesday, the company announced Gemini 1.5 Flash, the latest addition to the Gemini model series.

“We heard from developers that they want something faster and even more cost-effective,” said Demis Hassabis, CEO of Google DeepMind, in a press briefing.

The discovery comes as tech companies are increasingly refocusing their product development and deployment around generative AI, which is of particular importance to Google as the new tools give users more sophisticated and creative ways to access online information compared to the traditional web. search.

OpenAI on Monday released a new AI model and a desktop version of ChatGPT, along with a new user interface. The new model, the GPT-4o, is twice as fast as the GPT-4 Turbo and half the price, the company said.

Google also announced an improvement Gemini 1.5 Pro model that has the ability to make sense of multiple large documents — 1,500 pages in total — or summarize 100 emails, according to a vice president working on Gemini.

Gemini 1.5 Pro will soon be able to process an hour of video content or codebases with more than 30,000 lines, Hsiao said.

“You can quickly get answers and insights on full documents, such as understanding the details of the pet policy in your lease or comparing key arguments from multiple long research papers,” Hsiao said.

The latest OpenAI upgrade announced this week brings with it improved quality and speed of ChatGPT for 50 different languages. It will also be accessible through OpenAI’s application programming interface (API), allowing developers to start building apps using the new model immediately, executives said.

With 35 languages, Google says Gemini 1.5 Pro has a 2 million token window that measures context and shows how much information the model can process at once. The new model has improved local thinking, planning and image understanding, company executives said.

“It offers the longest context window of any underlying model to date,” Alphabet CEO Sundar Pichai said at the press briefing. At the event, he gave an example of a parent asking a Gemini to summarize all recent emails from their child’s school.

Gemini 1.5 Pro will initially be available for testing in Workspace Labs. Gemini 1.5 Flash will also be available for testing in Vertex AI, which is Google’s machine learning platform that allows developers to train and deploy AI applications.

WATCHING: AI dominates the annual list of disruptors