Google has just made its most advanced generative AI model, Gemini, available for businesses to use in app development. Gemini comes in three versions: Ultra, Pro, and Nano. The Pro version is now accessible through an API and can be used for free, although there are some usage limits.
In addition to Gemini, Google announced several other AI advancements, including an upgraded text-to-image tool called Imagen 2 and a set of models tailored for the healthcare industry.
Developers can access Gemini Pro through the Google AI Studio, which provides a web-based platform for creating prompts and obtaining an API key for app development. This version includes a 32K context window for text generation, which Google plans to expand in the future. There’s also a new Gemini Pro Vision endpoint that accepts both text and images as input and provides text output.
Sundar Pichai highlighted that the Gemini API offers a wide range of features, such as function calling, embeddings, semantic retrieval, custom knowledge grounding, and chat functionality. It supports 38 languages and is available in over 180 countries. Gemini Pro will also be integrated into Vertex AI, Google Cloud’s comprehensive AI platform, which offers tools, managed infrastructure, and built-in privacy and safety features.
Google aims to gather feedback from developers to refine Gemini Pro and prepare for the launch of the more advanced Gemini Ultra next year, designed for more complex tasks.
Currently, Gemini Pro and Gemini Pro Vision can be used for free, with a limit of 60 requests per minute. This free access will continue until the models become generally available next year. After that, Google plans to charge based on the number of characters or images processed. The pricing is set at $0.00025 per 1,000 characters and $0.0025 per image for input, with the same rate of $0.0005 per 1,000 characters for output.
Some users have noted that Google’s pricing is higher than competitors like OpenAI, as Google charges per character rather than per token, which can represent entire words.
In addition to Gemini Pro, Google has updated Vertex AI with Imagen 2, which can generate a variety of creative and realistic logos, emblems, and lettermarks. It also performs better in rendering text in multiple languages. Google is also making MedLM, a set of models fine-tuned for the healthcare industry, available to US-based organizations through Vertex AI. This builds on the Med-PaLM 2 model introduced earlier this year and will soon receive a Gemini-based upgrade.