AI21, an enterprise AI systems provider, has launched Jamba, a model that integrates the Mamba Structured State Space model (SSM) technology and transformer components.
The model is a production-grade Mamba-based model claiming to provide 3x more throughput on lengthy contexts compared to Mixtral 8x7B. Additionally, the model is claimed to accommodate up to 140K context on a single GPU and provides access to a 256K context window.
The model will be available on Hugging Face and NVIDIA API in the future.
By using this site, you agree to allow SPEEDA Edge and our partners to use cookies for analytics and personalization. Visit our privacy policy for more information about our data collection practices.