AI21 Labs is an Israeli technology company specializing in natural language processing (NLP) and Generative AI. The company develops AI systems that can understand and generate natural language, with applications ranging from writing assistance to enterprise-scale language models. Founded in Tel Aviv in 2017, AI21 Labs has positioned itself as a significant player in the competitive field of Large language model development, with its Jurassic series and Jamba models targeting both consumer and business markets.
The company's product portfolio includes Wordtune, an AI-based writing assistant that suggests paraphrases and rewrites, and AI21 Studio, a platform for building custom AI applications. AI21 Labs has secured substantial venture capital funding from investors including Google and Nvidia, and has formed strategic partnerships with cloud providers to distribute its models.
History
AI21 Labs was founded in November 2017 by Yoav Shoham, Ori Goshen, and Amnon Shashua in Tel Aviv, Israel. The company raised $9.5 million in seed funding in January 2019. On October 27, 2020, it launched Wordtune, an AI-powered writing assistant that understands context and suggests textual improvements. Google named Wordtune one of its favorite extensions of 2021.
In August 2021, the company introduced AI21 Studio and Jurassic-1, a natural language processing system with a token vocabulary exceeding 250,000. Financially, the company closed a $20 million investment from Walden Catalyst in November 2021, followed by a $25 million Series A round led by Pitango First later that month. In July 2022, AI21 Labs raised $64 million in a Series B round led by Ahren, with participation from Shashua, Walden Catalyst, Pitango, TPY Capital, and Mark Leslie.
In January 2023, the company launched Wordtune Spices, a Generative AI tool that generates text options to enhance sentences. On March 9, 2023, it released Jurassic-2, claiming improved response times, broader language support, and advanced instruction following. A $155 million Series C round closed on August 31, 2023, with new investors including Google and Nvidia.
Product Evolution and Model Releases
In March 2024, AI21 Labs released Jamba, an open-weights Large language model built on a hybrid architecture combining Mamba state-space models with Transformer (architecture) components using mixture of experts, supporting context lengths up to 256,000 tokens. The Jamba 1.5 family followed, and in September 2024 the company partnered with Amazon Web Services to make these models available on Amazon Bedrock for enterprise deployment.
In March 2025, AI21 Labs released Jamba 1.6 for private enterprise use, claiming it outperformed other open models across multiple benchmarks. The same month, the company launched Maestro, an AI planning and orchestration system designed to improve accuracy of models like GPT-4o and Claude 3.5 Sonnet on complex tasks.
Management and Advisors
AI21 Labs operates under a co-CEO structure. Amnon Shashua serves as chairman and co-founder, while Ori Goshen and Yoav Shoham act as co-CEOs and co-founders. The company maintains a board of academic advisors from leading institutions, including Dan Jurafsky from Stanford AI Lab, Christopher Ré from Stanford University, Sebastian Thrun from Stanford University, and professors from the Hebrew University of Jerusalem such as Omri Abend, Yonatan Belinkov, and Shai Shalev-Shwartz. Additional advisors include Kevin Leyton-Brown from the University of British Columbia, Dan Roth from the University of Pennsylvania, and Nick McKeown from Stanford University.
Technology and Partnerships
The company's models leverage advanced Deep learning techniques, including mixture of experts architectures and hybrid convolutional approaches. Its partnership with Amazon Web Services via Amazon Bedrock enables integration of Jamba models into enterprise applications, and the company has also collaborated with cloud providers like Google Cloud and Oracle Cloud Infrastructure for model distribution. AI21 Labs competes in the same market as OpenAI, Anthropic, and Google DeepMind, but differentiates through open-weights models like Jamba and a focus on Hebrew and multilingual NLP capabilities.
See Also
- chees-computer placeholder
- OpenAI development ecosystem