Jamba
7 models · AI21 Labs
AI21's hybrid Mamba-Transformer family (1.5–1.7 Large/Mini, open Jamba Reasoning 3B) — SSM hybrids offering 256K context with much lower memory than pure attention.
7 models · AI21 Labs
AI21's hybrid Mamba-Transformer family (1.5–1.7 Large/Mini, open Jamba Reasoning 3B) — SSM hybrids offering 256K context with much lower memory than pure attention.