Provider pricing

AI21 Labs AI API pricing

AI21 Labs builds the Jamba family — hybrid Mamba-Transformer models that combine Transformer quality with Mamba's linear-time long-context efficiency. Jamba 1.5 Large and Mini both ship with 256K context windows and strong RAG, function calling, and structured-output support.

Founded2017
HeadquartersTel Aviv, Israel
Tracked models3

Best for

Long-document RAG, contract analysis, and workloads where 100K+ context is the norm.

Strengths

  • Hybrid Mamba-Transformer: linear-time inference on long contexts
  • 256K context window across the lineup
  • Strong grounded generation and citation support for RAG
  • Available on AI21 Studio, Bedrock, Vertex, and Azure

Tracked AI21 Labs models

ModelInput $/MOutput $/MContextTierBest for
Jamba 1.7 Large$2.00$8.00256,000FlagshipHybrid Mamba-Transformer flagship with 256K context
Jamba 1.7 Mini$0.200$0.400256,000BudgetCost-efficient hybrid SSM for long-context RAG
Jamba Instruct$0.500$0.700256,000Mid-tierInstruction-tuned variant for enterprise deployments