Google Gemini Pro: Launching Soon for Businesses and Developers

Google is unveiling its latest innovation, the Gemini model, offering businesses and developers the first look at its powerful large language capabilities through an accessible API. Gemini is available in three sizes: Ultra, Pro, and Nano. From today, developers can access the Gemini Pro API via Google’s free web-based developer tool, AI Studio (formerly known as Makersuite), while enterprises can integrate it through Google Cloud’s Vertex AI platform, enabling them to create applications promptly.

Google has announced plans to refine Gemini Pro further based on user feedback in the upcoming weeks. “We eagerly anticipate the innovative applications that developers and enterprises will create with Gemini,” the company shared in a recent blog post. Currently, Gemini Pro powers Bard, Google’s conversational AI designed to compete with ChatGPT. One key feature of the initial version is its 32,000 token context window, which can process approximately 5,333 words. In comparison, OpenAI’s GPT-4 Turbo can handle up to 128,000 tokens. However, future iterations of Gemini Pro are expected to significantly expand this capacity.

Among Gemini Pro’s features are support for 38 languages, function calling, embeddings, semantic retrieval, and custom knowledge grounding. At present, the API operates with text input and output exclusively. However, a multimodal endpoint—Gemini Pro Vision—has been launched to accept both text and visual inputs, such as images and videos, generating text outputs based on them.

Currently, the Gemini Pro API is free to use, but it is limited to a maximum of 60 queries per minute. A pay-as-you-go version is set to be introduced soon, promising fewer restrictions with a pricing structure that Google describes as "competitively priced." The pricing for Gemini Pro has been established at $0.00025 per thousand characters and $0.0025 per image, while output is charged at $0.0005 per thousand characters. Inputs and outputs from the free version will be utilized by Google to enhance its offerings, while data from the paid version will remain private.

In addition to Gemini Pro, Google is expanding its Vertex platform with new models, including Imagen 2, the latest AI image generation model from Google DeepMind. This advanced text-to-image diffusion model can produce high-quality images and even realistic logos for businesses. Furthermore, it can render text in multiple languages.

Another significant addition is MedLM, a suite of foundation models fine-tuned specifically for the healthcare sector. Built on the Med-PaLM 2 model, MedLM is intended for applications such as medical notetaking and answering healthcare-related questions. Currently, this model is exclusively accessible to U.S.-based Vertex users, with future plans to broaden its availability in the weeks ahead. Google also aims to incorporate Gemini-based models into the MedLM suite shortly.

Finally, the Duet AI for Developers tool is now generally available. This collaboration tool helps developers streamline their application-building process and can be integrated into various Google Cloud interfaces for code generation and chat assistance. Over the next few weeks, Gemini will be integrated into Duet AI, which is also expanding into security operations, enhancing collaboration for defenders within a unified SecOps platform.

With these innovative tools, Google is setting the stage for a new era of AI-driven applications that promise to enhance productivity, creativity, and security across industries.