Google improves Gemini 1.5 Pro and introduces the new Gemini 1.5 Flash.

At the Google I/O 2024 event, AI was the common thread of a number of new updates and developments. For example, the company is launching a new AI model Gemini 1.5 Flash and making quality improvements to its latest Gemini 1.5 Pro model in areas such as translation, coding, reasoning and more. Users with Gemini Advanced can also take advantage of a context window with 2 million tokens. The new updates are now available in the model.

Improvements Gemini 1.5 Pro

In February, Gemini 1.5 Pro was launched, the latest version of the Gemini series, capable of accurately processing up to millions of tokens. Just three months later, at the Google I/O event on May 14, Google is bringing new quality improvements to the model, such as translation, encoding, reasoning, and more. The company said the latest version achieved improvements in several benchmarks, including MMMU, MathVista, ChartQA, DocVQA, InfographicVQA and more.

Additionally, Gemini 1.5 Pro will be made available to consumers in Gemini Advanced. This allows users to get AI assistance with large volumes of work such as long PDFs. The updates make the model on Android better able to apply AI to what appears on your screen. Gemini 1.5 Pro features a 1 million token context window and allows users to use text, images, audio and video as input.

To access 1.5 Pro with a 2 million token context window, users must waitlist in Google AI Studio or Vertex AI for Google Cloud customers.

AI model updates

In addition to the Gemini models, Gemma is also getting upgrades with the launch of Gemma 2 in June. This next generation is optimized for TPUs and GPUs and will launch with 27B parameters. In addition to Gemma 2B and Gemma 7B, the Gemma family has another sister, PaliGemma. That’s the first Vision language model from Google.

In addition, Gemini Nano, the model that runs on smartphones, also benefits from extensions and accepts images in addition to text. Applications using Gemini Nano with Pixel Multimodality can understand images, sounds and spoken language.

Gemini 1.5 Flash

A Google I/O event with AI as the central theme is not possible without the announcement of a new AI model. Although Google recently launched its latest model Gemini 1.5 Pro, the next model is already ready: Gemini 1.5 Flash. This smaller Gemini model is optimized for high-frequency tasks where speed and model response time play an important role.

It is also the fastest Gemini model offered in the API and offers a more cost-effective alternative than Gemini 1.5 Pro. Gemini 1.5 Flash is available today in preview on Google’s AI Studio and Vertex AI.

Android and Workspace integration

Gemini will also replace Google Assistant and become the default AI assistant on Android phones by long pressing the power button. Gemini can be used across different services and apps and offers multi-modal support when needed.

Google Workspace is also getting AI upgrades and the Gemini Pages panel of Gmail, Docs, Drive, Slides and Sheets is being replaced with Gemini 1.5 Pro. This offers many improvements to users as this model provides a longer context window and extended reasoning.

A complete overview of all new announcements for the Google I/O 2024 event can be found here.