Jamba Large

Jamba Large is AI21 Labs' hybrid SSM-Transformer long-context model, built for enterprise agents and grounded generation. It is the first version of the jamba line in the registry, released on 2025-07-01. It has a context window of 256000 tokens and a maximum output of 4096 tokens. Pricing starts at $2.00 per million input tokens and $8.00 per million output tokens. The weights are open. It supports tool calling, accepts and returns text only, and does not have a reasoning mode. Its knowledge cutoff is 2024-08-22.

First version of its line in the registry.

• ·256K token context window for long documents
• ·Hybrid SSM-Transformer architecture for efficiency
• ·Tool calling for enterprise agent workflows
• ·Open weights for self-hosting and customization
• ·Grounded generation for factual responses

• Reach for it for long-context enterprise document analysis
• Reach for it for building agents that call external tools
• Reach for it for grounded generation over large corpora
• Reach for it for text-only tasks with high token volumes

1 host serve this model at the price above · the maker's documentation

The registry keeps the six most recent versions of a line. Older ones stay in the file and can be surfaced without a rebuild.

What is the context window and output limit?

The context window is 256000 tokens and the maximum output is 4096 tokens per request.

Is the model open weights and what does it cost?

Yes, the weights are open. Pricing starts at $2.00 per million input tokens and $8.00 per million output tokens.

Does it support tool calling or reasoning?

It supports tool calling, but it does not have a reasoning mode. It accepts and returns text only.