Gemini 1.0 Pro is now generally available to all Vertex AI customers, while Gemini 1.0 Ultra is available to Vertex AI customers via an allowlist.
Google recently switched entirely to Gemini for its model naming, further eliminating Bard. There are three Gemini versions for enterprises: Nano, Pro and Ultra. Nano is optimized for on-device tasks, while the more powerful Pro and Ultra offer more functionality. Gemini Pro is the “best model for scaling across a wide range of tasks.” Gemini Ultra is specifically designed for highly complex tasks and is Google’s largest model.
Role Vertex AI
Extending access to the models covers Gemini 1.0 Pro and Gemini 1.0 Ultra. If companies use Vertex AI, Google’s offering of tools for building and using generative AI, they can now deploy recent versions of the Gemini models.
The Gemini API in Vertex AI allows developers to build AI agents and apps. Vertex AI will enable developers to customize Gemini models to meet business needs. There is support for Low-Rank Adaptation, allowing developers to adapt the model more efficiently and cost-effectively. In the coming months, the Gemini API will also gain access to the personalization technique reinforcement learning from human feedback.
Google aims to improve the response of the Gemini model further by adding the ability for the model to use current information and business data. Function calls allow developers to connect the model to external APIs for transactions and other actions.
Finally, Google is taking steps to manage and scale the Gemini model into production. To this end, the tech giant offers tools that ensure that applications, once developed, can be easily deployed and maintained. Vertex AI offers an automated evaluation tool for generative AI models called Automatic Side by Side. This feature compares the responses of models using a standard set of criteria. This helps developers gain insight into Gemini’s performance and adjust prompts and tuning based on the feedback.
Tip: Google’s more powerful AI chatbot is available for 22 euros a month