AI21 Labs

AI21 Labs, established in 2017, focuses on creating Foundation Models and AI Systems tailored for enterprise use. The company’s mission is to bridge the gap between cutting-edge research and enterprise-ready AI systems. AI21 Labs began its journey with Wordtune, an AI writing companion now used by over 10 million people globally. The company offers a variety of specialized Task-Specific Models (TSMs) designed to address high-impact use cases efficiently and cost-effectively, with capabilities such as paraphrasing, text improvements, grammatical error corrections, summarization, and text segmentation. AI21 Labs’ Jamba model, a hybrid SSM-Transformer model, is notable for its efficiency, delivering 3X throughput on long contexts and fitting 140K context on a single GPU. Jamba-Instruct, an instruction-tuned version of Jamba, is available in public preview. The company’s RAG Engine provides Retrieval-Augmented Generation alongside its Task-Specific and Foundation Models. AI21 Labs’ models are accessible on platforms like NVIDIA’s API Catalog, Google Cloud Marketplace, and Amazon’s Bedrock. The company recently completed a $208 million oversubscribed Series C funding round. AI21 Labs’ Summarize API has been evaluated to outperform OpenAI’s models in human assessments, and its Jurassic-2 models are highly ranked for accuracy and reliability by Stanford’s HELM. The company offers flexible deployment options, including SaaS, APIs on the AI21 Platform, Cloud, Private, VPC, or On-Prem. AI21 Labs’ Contextual Answers API provides answers based entirely on the provided context of documents, ensuring precise and relevant responses.