Exploring Vertex AI's Latest LLM Models: Gemini 1.5 Pro and More

Vertex AI continues to push the boundaries of artificial intelligence with its latest updates and large language models (LLMs). In this post, we'll explore the recent enhancements, focusing on the Gemini 1.5 Pro model and other notable additions.

Gemini 1.5 Pro Model

The Gemini 1.5 Pro model stands out as a significant upgrade within the Gemini family. This model excels in long-context understanding across various modalities, featuring a groundbreaking context window of up to 1 million tokens. Such capacity enables it to handle vast amounts of data effectively.

Key features include:

• Support for text, images, videos, and audio processing
• Capabilities like transcription and cross-modal analysis
• Available in private preview via AI Studio and Vertex AI

The Gemini 1.5 Pro is set for a wider release with improved latency, making it a versatile tool for diverse AI tasks.

Other Recent Models and Updates

Vertex AI has also introduced other models and updates to enhance its offerings:

• Llama 3.1 405B Model: Available in preview, this model supports synthetic data generation, model distillation, steerability, math, tool use, and multilingual translation.
• Anthropic Claude 3.0 Opus Model: Generally available and accessible through Vertex AI.
• Gemini 1.5 Flash and Pro Models: These models offer enhanced performance and context handling and are generally available.
• New Models in Model Garden: Includes Hugging Face embedding and PyTorch models, as well as multimodal models like MaMMUT.

Enhanced Features and Capabilities

Vertex AI is not just about new models; it also provides robust features for fine-tuning and deployment:

• Fine-Tuning with LoRA: Efficiently updates models by introducing smaller matrices instead of retraining the entire model.
• Agent Builder: Simplifies the creation of virtual agents using LLMs, with tools for grounding outputs in Google Search and other data sources.
• MLops and Prompt Management: Expanded capabilities to help enterprises manage and optimize their LLMs.

Regional Availability and Data Residency

Generative AI on Vertex AI is now available in multiple regions, including us-east5, me-central1, and me-central2. Additionally, data residency for models like Gemini, Imagen, and Embeddings API has expanded to 11 new countries, ensuring broader accessibility and compliance with local data regulations.

These updates underscore Vertex AI's commitment to advancing AI capabilities and expanding its global reach. Stay tuned for more exciting developments!