Preeti
Jamba is a groundbreaking Large Language Model (LLM) that combines Transformer, Mamba, and mixture-of-experts (MoE) layers to address the limitations of traditional models. With a context window of 256K, Jamba outperforms other models in its size class, delivering three times the throughput on long contexts. It can accommodate up to 140K contexts on a single GPU, making it highly scalable and efficient for large-scale language tasks. Jamba’s release marks a milestone in LLM innovation by integrating Mamba alongside the Transformer architecture, offering unprecedented efficiency, throughput, and scalability.
AI21, the company behind Jamba, is a leader in generative AI and LLMs, with a mission to empower businesses with cutting-edge AI systems. They have raised $336 million in funding and offer enterprise solutions and consumer applications. Jamba’s integration with the NVIDIA API catalog as a NIM inference microservice streamlines its accessibility for enterprise applications. The release of Jamba with open weights under the Apache 2.0 license encourages collaboration and innovation in the open-source community.
AI21’s consumer product, Wordtune, is a popular AI-based reading and writing assistant that understands context and meaning, helping users refine and generate text quickly. With the developer platform AI21 Studio, businesses can build their own generative AI-driven applications and services. Overall, Jamba represents a significant advancement in LLM technology, offering improved efficiency, throughput, and scalability for a wide range of language tasks.
Source link
Source link: https://www.prnewswire.com/news-releases/unveiling-jamba-ai21s-groundbreaking-hybrid-ssm-transformer-open-source-model-302102779.html
GIPHY App Key not set. Please check settings