NEW: You can now run Mamba-based architectures (shout-out to @tri_dao & @_albertgu for introducing Mamba), as well as hybrid Transformer–Mamba models (like our own #Jamba), with the enhanced vLLM v1 engine. Kudos to our very own @AsafGardin and Amir Koblyansky for their

→ View original post on X — @ai21labs,