Google Makes Gemini Pro Generally Available on Vertex AI
Google's Gemini Pro model is now publicly available on its Vertex AI platform. The move provides enterprise customers with a managed, scalable way to access Gemini's advanced reasoning and multi-modal capabilities. This positions Vertex AI as a more direct competitor to Azure's OpenAI service for building enterprise applications like RAG and vector search pipelines.
- The general availability on February 15, 2024, included support for adapter-based tuning like Low-Rank Adaptation (LoRA), allowing for more efficient customization of the model for specific enterprise domains. Vertex AI also supports other tuning methods like distillation and reinforcement learning from human feedback (RLHF). - Initial pricing for Gemini Pro on Vertex AI was set at $0.00025 per 1,000 characters for input and $0.0005 per 1,000 characters for output, making it a cost-competitive option against models like OpenAI's GPT-3.5 Turbo. - The first version of Gemini Pro came with a 32K token context window for text. This has since been significantly expanded in subsequent models like Gemini 1.5 Pro, which offers a 1 million token context window, enabling the processing of large documents, codebases, or even hours of video. - For multimodal applications, a dedicated Gemini Pro Vision endpoint was made available from the start, capable of processing both text and images to generate text outputs. This allows for the development of applications that can understand and analyze visual data. - Early enterprise adoption before the general release included companies like Samsung, which utilized the model for summarization features in its native applications, and Palo Alto Networks, which tested Gemini for creating intelligent product agents. - The Vertex AI platform provides a suite of MLOps tools and a unified environment that supports not just Gemini, but over 130 other models from Google and third parties, facilitating an end-to-end workflow from model customization to deployment. - Developers can interact with Gemini Pro through SDKs for various programming languages, including Python, Node.js, and Swift. The platform also offers features for grounding models with a company's own data to enhance response accuracy and reduce hallucinations. - Gemini Enterprise, a business-focused offering, integrates with Vertex AI to provide enhanced security and governance features such as customer-managed encryption keys, VPC Service Controls, and data residency options to meet strict compliance requirements.
