NVIDIA

NVIDIA Nemotron

NVIDIA's models, often hybrid (Mamba + attention): a much lighter context cache.

Versions