Google has introduced Gemini Pro, the latest addition to its GenAI model family
In a recent development, Google has introduced Gemini Pro, the latest addition to its GenAI model family, to Vertex AI, the company’s fully managed AI development platform. Following its recent debut on Bard and the Pixel 8 Pro, Gemini Pro is now available in public preview, offering Google Cloud customers access to its capabilities via the new Gemini Pro API.
Gemini Pro is a lightweight version of the more advanced Gemini Ultra, currently in private preview for a select set of customers. The API is free to use within certain limits and supports 38 languages and regions, including Europe. Notable features include chat functionality, filtering, and a state-of-the-art natively multimodal model that boasts sophisticated reasoning and advanced coding skills.
The Gemini Pro API, integrated into Vertex AI, accepts text as input and generates text as output. Additionally, a new endpoint called Gemini Pro Vision, also in preview, enables the processing of both text and imagery, including photos and videos, outputting results akin to OpenAI’s GPT-4 with Vision model.
Developers within Vertex AI can customize Gemini Pro to specific contexts and use cases using fine-tuning tools similar to those available for other models hosted on Vertex, such as Google’s PaLM 2. Gemini Pro can be connected to external APIs to perform specific actions and grounded to enhance the accuracy of responses by leveraging data from various sources.
Citation checking, an existing Vertex AI capability, now supports Gemini Pro, providing an additional fact-checking measure by highlighting the sources of information used by the model to generate responses.
Addressing concerns about Gemini’s image analysis performance, Google has introduced Gemini Pro Vision to process images, offering users the opportunity to evaluate the model’s image comprehension capabilities
Gemini Pro comes with control, moderation, and governance options, emphasizing its strengths and pushing back against any perception of it being less powerful. To attract developers, Google is offering attractive pricing, with input for Gemini Pro priced at $0.00025 per character and output at $0.00005 per character, reduced from the pricing of its predecessor. Moreover, Gemini Pro is free to try for Vertex AI customers until early next year.
Google is also enhancing Vertex AI with new features, including the integration of Gemini Pro into custom-built conversational voice and chat agents, as well as search summarization, recommendation, and answer generation features. These additions are expected to roll out in early 2024.
Other updates to Vertex AI include Automatic Side by Side (Auto SxS), a response to AWS’ Model Evaluation on Bedrock, allowing developers to evaluate models in an on-demand, automated fashion. Google is expanding Vertex AI with models from third parties, introducing step-by-step distillation, and extending its indemnification policy to cover outputs from PaLM 2 and Imagen models.
While Google currently does not extend its Vertex AI indemnification policy to cover customers using the Gemini Pro API, the company has stated its intention to do so once the API launches publicly.
