AI21 Begins Offering Jamba-Instruct on Microsoft Azure AI

AI21's latest foundation model, Jamba-Instruct, is now available through Microsoft Azure AI's Models-as-a-Service (MaaS). Through this partnership, enterprise developers can leverage a cutting-edge language model without the burden of managing cloud infrastructure.

Jamba-Instruct adopts the world's first production-grade SSM-Transformer hybrid architecture, securing performance, quality, and cost efficiency all at once. Its key features are as follows:

• 70K context window: Effectively handles long contexts, making it optimized for document understanding and building advanced RAG engines.
• Cost-efficient processing: Thanks to its innovative hybrid structure, it can process long contexts with fewer cloud resources, reducing computing costs.
• Best cost-performance: It matches or exceeds competing models on major benchmarks while offering highly competitive pricing.

Jamba-Instruct is an instruction-tuned model equipped with safety guardrails, chat capabilities, and complex instruction comprehension, making it immediately applicable to enterprise applications. It also securely protects enterprise data through Azure AI's security platform and can be flexibly used via a pay-as-you-go API.