Jamba Mamba-based Model for Superior AI Performance

Jamba Mamba-based model represents a groundbreaking advancement in language model technology, combining Mamba Structured State Space (SSM) with traditional Transformer architecture to deliver a 256K context window and superior performance. Licensed under Apache 2.0, Jamba offers remarkable throughput and efficiency, fitting up to 140K context on a single GPU, and is available on Hugging Face and NVIDIA API catalog.

AI21labs
Not Applicable
March 29, 2024
Introducing Jamba: AI21's Groundbreaking SSM-Transformer Model
Jamba v0.1 on HuggingFace