Two startups developing large language models have raised substantial new funding, backing different approaches to building more capable AI systems. Imbue AI is working toward general-purpose agents that can make robust decisions, while AI21 Labs focuses on large-scale language models combined with what it calls “neurosymbolic systems.”
Imbue starts with agents for code
Imbue AI, formerly Generally Intelligent, raised $200 million in a Series B funding round. Its investors include Astera Institute, Nvidia, Cruise CEO Kyle Vogt, and Notion co-founder Simon Last. The company has raised $220 million since inception and is valued at $1 billion.
The company’s stated aim is to create everyday AI agents with robust decision-making capabilities. Its plan is to develop language models designed to reason, beginning with agents for code generation.
Code generation is a starting point, rather than the final product Imbue describes. The company’s underlying assumption is that working with code can promote reasoning and help agents interact effectively with computers. On that view, code agents provide a foundation for building more powerful and general AI agents.
A full-stack approach to AI agents
Imbue says its work spans several connected areas. It plans to train foundational models, build experimental agents and interfaces, develop tools and infrastructure, and study the theoretical basis of how models learn.
That combination reflects the breadth of the company’s goal. A capable agent depends not only on a language model, but also on how the model is trained, the tools it can use, and the interface through which people work with it. Imbue’s plans bring these pieces together as part of a longer effort to make agents that can reason and act.
The company’s focus on code generation therefore has a broader purpose in its strategy. If code work helps develop reasoning and computer interaction, it can serve as groundwork for agents that operate beyond coding tasks.
AI21 Labs raises $155 million
AI21 Labs announced the close of a $155 million Series C funding round, valuing the company at $1.4 billion. Investors include Google and Nvidia, along with Walden Catalyst, Pitango, SCB10X, b2venture, Samsung Next and Prof. Amnon Shashua. The company has raised $283 million in total.
Founded in 2017 by Amnon Shashua, Ori Goshen, and Yoav Shoham, AI21 Labs develops large-scale language models that understand and generate language, combined with “neurosymbolic systems.” The company’s approach links language capabilities with systems that incorporate symbolic methods.
AI21 Labs released its Jurassic-1 Jumbo language model, with 178 billion parameters, in August 2021. Customers can access its models through AI21 Studio or an API. The services include chatbots, summaries, and ad copy, and the company targets enterprise customers.
Different routes, a shared ambition
The funding rounds support companies with distinct plans for language model development. Imbue is emphasizing reasoning-oriented models and agents, with code generation as an initial step. AI21 Labs is building and offering large-scale language models for enterprise uses, pairing language understanding and generation with neurosymbolic systems.
In both cases, the new capital accompanies an effort to move beyond language generation alone. Imbue’s plans center on agents that can make decisions and interact with computers. AI21 Labs’ model access is tied to practical services such as chatbots and summarization. Their strategies differ, but each places language models within a wider ambition for useful AI capabilities.