Imbue, a San Francisco-based AI startup, has raised $200 million in a Series B round at a valuation of over $1 billion, it announced on Thursday. The investors in the round include Astera Institute, NVIDIA, Cruise CEO Kyle Vogt, and Notion co-founder Simon Last, and others. The valuation makes it one of the highest valued AI startups in the world. Not in the top ten though.
The round takes total financing of the startup to $220 million. Formerly known as Generally Intelligent, the startup had previously raised $20 million from Astera Institute, Anthropic’s co-founder Tom Brown, former tech lead of OpenAI robotics Jonas Schneider, and Dropbox co-founders Dropbox co-founders Drew Houston and Arash Ferdowsi.
Founded in 2021 by Kanjun Qiu and Josh Albrecht, Imbue wants to build ‘AI systems’ that can reason and code. The company said that it is currently applying its models to develop agents that it finds useful internally, starting with agents that can code, “Ultimately, we hope to release systems that enable anyone to build robust, custom AI agents that put the productive power of AI at everyone’s fingertips.”
The AI startup argues that the current AI systems are very limited in their ability to complete even simple tasks for users because of their inability to reason, “Robust reasoning is necessary for effective action. It involves the ability to deal with uncertainty, to know when to change our approach, to ask questions and gather new information, to play out scenarios and make decisions, to make and discard hypotheses, and generally to deal with the complicated, hard-to-predict nature of the real world.”
That’s why it builds foundation models, which it asserts are tailored for reasoning. It achieves this by crafting pre-training data specifically designed to strengthen good reasoning patterns and by developing techniques that allocate significantly more computational resources during inference to reach robust conclusions and execute actions.
The startup hopes that its AI agents will help free people up to focus on things they care about, “When we build AI agents, we’re actually building computers that can understand our goals, communicate proactively, and work for us in the background. Today, we’re glued to our screens because our computers need to be micromanaged—if we’re not in front of the screen, very little gets done. Truly useful AI agents will fundamentally change this.”
This is perhaps a more sophisticated way of saying that they want to build AGI.
The company, like some other AI labs, including OpenAI, Cohere, and Inflection, uses a full-stack approach for building this; which includes training foundation models, prototyping experimental agents and interfaces, building robust tools and infrastructure, and understanding the theoretical underpinnings of how models learn.
Imbue is training its models with over 100 billion parameters, which have been optimized to excel on its internal reasoning benchmarks, using a computer cluster equipped with over 10,000 Nvidia H100 GPUs.