# Hybrid SSM-Transformer architecture

AI21 Jamba 1.5 is a series of advanced foundation models with powerful long context processing capabilities and efficient inference speed, suitable for various business scenarios.

AI21 Jamba 1.5 Mini is an advanced hybrid SSM-Transformer instruction-following foundation model with efficient long-context processing capabilities and fast inference speed.

Jamba is a state-of-the-art hybrid SSM-Transformer large language model that combines the advantages of Mamba architecture with Transformer, supporting 256K context length, surpassing models of similar scale in throughput and performance.