Google's Evolved Multimodal AI Model Gemini Pro Unfolded for Development Purposes

Recently, tech giant Google unraveled a series of innovative updates related to its newly launched multimodal AI model, Gemini. This came only a week after unveiling this cutting-edge technology which now promises to bolster developers' creative works.

At the time of its unveiling, Google highlighted the availability of Gemini in three distinct versions — Ultra, Pro, and Nano. These are tailored according to different size and complexity requirements. Of note, Gemini Nano has already been incorporated into Android with the Pixel 8 Pro. Furthermore, Google Bard now benefits from an adapted version of Gemini Pro.

Breaking new ground, Google has now announced that developers can access the wide array of features Gemini Pro offers through the Gemini API. Developers are afforded the use of functional calling, embeddings, semantic retrieval, custom knowledge grounding, and chat functionality.

Developers are also introduced to two main ways of working with Gemini Pro: Google AI Studio and Vertex AI on Google Cloud. Deemed as an easy starting point, Google AI Studio offers a web-based development tool, a free quota for 60 requests per minute, and quick-start templates to ease developers into the model.

Touted as a more sophisticated option, Google's Vertex AI on Google Cloud is a machine learning platform that permits full customization of Gemini. Loaded with other practical features such as full data control and integration with other Google Cloud modules, developers can comply with security, safety, privacy, and governance needs.

Currently, usage of Gemini within Vertex AI is free within the same rate as Google AI Studio's free quota. Nonetheless, this provision extends only until it hits general availability, projected for the following year. After which, charges will be at $0.00025 for 1000 characters and $0.0025 per image.

Google elaborates on the sophisticated capabilities of the Vertex AI environment including the ability to enhance Gemini with company data and develop search and conversational agents in a low-code environment akin to what powerful platforms such as AppMaster provide.

Currently, Gemini Pro receives and outputs text. But for developers eager to experiment with images, a dedicated Gemini Pro Vision endpoint accepts images alongside text and outputs text.

Google further fuels anticipation with a forecast to launch Gemini Ultra early next year. This is a larger model targeting more complex tasks. Google is also considering integrating Gemini with the Chrome and Firebase developer platforms.

Today also marks unveiling the upgraded version of the tech behemoth’s image-generation model, Imagen 2. This is now open for all Vertex AI users included in Google’s allowlist.

This advanced iteration of Imagen enables the production of aesthetically engaging, high-quality, and true-to-life images steered by natural language prompts. Its fascinating new features consist of text rendering to create text overlays on images, logo generation, and visual question and answers for caption generation.