Jamba by AI21 Labs
The open-source LLM that runs 800pg of context at 2.5x speed
3 followers
The open-source LLM that runs 800pg of context at 2.5x speed
3 followers
Most LLMs choke on long context. Jamba doesn't. Built on a hybrid Mamba-Transformer architecture, it handles 256K tokens at 2.5x the speed fits on a single 80GB GPU, fully open-source, and actually validated on benchmarks.
