Active
AI21 Labs
Language models, knowledge agents, and an AI gateway for application workloads.
- Jamba models for long-document applications
- Maestro agents for searching files and the web
- Gateway routing, caching, and context compaction
AI21 combines model development with tools for building and optimizing AI applications. Jamba is its model family; Maestro adds document retrieval and agent workflows. The Gateway sits between an application and its models to adjust routing, caching, and the amount of context sent on each turn.
How pricing works
AI21 model API usage is metered by tokens, with input and output priced separately. Cloud-hosted deployments follow the selected cloud provider's billing. The Gateway is currently free during its beta preview; that does not make the underlying model calls free.