Now Available:
Gemini 1.5 Flash: In public preview, Gemini 1.5 Flash offers a groundbreaking context window of 1 million tokens, designed to serve tasks like chat applications with speed and scale more efficiently than 1.5 Pro.
PaliGemma: PaliGemma, found in the Vertex AI Model Garden, is the first visual language model in the Gemma open model family, ideal for tasks like image captioning and visual question answering.
Coming Soon:
Imagen 3: Imagen 3 is capable of producing incredibly detailed and photorealistic images, making it the highest-quality text-to-image generation model to date.
Gemma 2: Gemma 2, the next generation of the open model family designed for a wide range of AI developer use cases, utilizes the same technologies used to build Gemini.
New Features in Vertex AI:
Google announced new features to help customers optimize model performance, including context caching, controlled generation, and batch API.
Context Caching: Context caching allows customers to actively manage and reuse cached contextual data, significantly reducing processing costs.
Controlled Generation: Controlled generation allows customers to define Gemini model outputs according to specific formats or schemas, ensuring the format and syntax of model outputs.
Batch API: Batch API targets use cases like classification and sensitivity analysis, enabling the super-efficient sending of numerous text prompt requests, speeding up developer workflows and reducing costs.
Agent Builder: New Open Source Integrations
Vertex AI Agent Builder empowers developers to create and deploy AI experiences through a range of tools, from code-first open-source editing frameworks like LangChain to codeless consoles using natural language. Google introduced Firebase Genkit and LlamaIndex to strengthen Agent Builder on Vertex AI.
Firebase Genkit: Firebase’s Genkit simplifies the development, deployment, and monitoring of production-ready AI agents.
LlamaIndex on Vertex AI: LlamaIndex streamlines the production process from data ingestion and transformation to embedding, indexing, retrieval, and deployment, offering a simple, flexible, open-source data framework to connect custom data sources to productive models.
Grounding with Google Search
Google announced the general availability of Grounding with Google Search, enabling customers to ground their outputs in their private databases or specified “enterprise accuracy” sources. Additionally, Google expanded the scope of generated output compensation to include outputs grounded with Google Search in Productive AI compensated services.
Google aims to democratize AI innovation with Vertex AI and accelerate organizations’ AI deployments in production.”
