AI21 Studio

AI21 builds foundation models and AI systems for enterprises, and its Studio offers API and SDK access to its Jamba models. The current site focuses on helping teams control the cost and quality of their AI stack. Three offerings are described. The Intelligent Gateway is a drop in endpoint that cuts token waste by routing, caching and compacting agent traffic, set up with two lines of configuration. Harness Optimization searches configurations against a customer's own private evals to find the best strategy under a chosen cost ceiling and quality floor, and is marked coming soon. Dedicated Model tunes small open models to match frontier quality on a customer's defined workloads. The vendor says combining cheap open models with frontier models through routing and custom validators delivers frontier quality at about 30 percent of the cost, and it publishes research notes on the approach. The foundation model APIs include Jamba Mini, described as efficient and lightweight, and Jamba Large, described as a long context model. It is aimed at enterprise engineering and AI teams running agents in production. Pricing is usage based and covers unlimited seats. A free trial provides $10 in credits for 7 days with no credit card. Jamba Mini costs $0.2 per million input tokens and $0.4 per million output tokens. Jamba Large costs $2 per million input tokens and $8 per million output tokens. A Custom Plan adds volume discounts, premium rate limits, private cloud hosting, priority support, a dedicated account manager and consultancy, and is available through sales.