Nous Research Inc., the developer of the popular Hermes artificial intelligence agent, today announced that it has raised $90 million in funding.
Early-stage startup fund Robot Ventures led the Series B round. It was joined by Nvidia Corp., Samsung Electronics Co., Y Combinator and several others. Nous is now valued at $1.5 billion.
The company released Hermes in February under an open-source license. Nous says that the agent has since been downloaded more than 24 million times and boasts several thousand open-source contributors. It accounts for an estimated 2.5% of the world’s token consumption, a metric tied to AI models’ hardware use.
Hermes ships as a desktop app that can interact with the user’s files. A consumer could, for example, ask the agent to organize a collection of documents by category. Hermes can also access the programs on the user’s machine. It can use a browser to find web data, run code in a Docker container and perform related tasks.
Under the hood, Hermes uses popular large language models such as Claude to complete tasks for users. It also supports a lineup of LLMs developed by Nous. They’re fine-tuned, or customized, versions of open-source LLMs.
Nous’ latest collection of LLMs is called the Hermes 4 series. The company customized the models using a training dataset that comprised 5 million synthetically generated records. According to Nous, about two thirds of the records were specifically created to enhance Hermes 4’s reasoning capabilities. The company generated the data using a custom tool called DataForge.
A so-called router enables Hermes to send each user prompt to the LLM best equipped to process it. The mechanism takes into account factors such as inference costs, speed and output quality.
Another built-in module personalizes LLM responses based on user preferences. Optionally, users can enhance Hermes’ personalization features by connecting it to external databases such as the agent-optimized RetainDB relational store. According to Nous , adding an external database enables Hermes to retain preferences and project data across user sessions.
The company monetizes its agent with three paid editions. The first, Hermes Cloud, enables consumers to launch agents in a few seconds without setting up any infrastructure. Nous’ two other services are geared towards organizations. Hermes Business and Hermes Enterprise enable administrators to set team-level token use limits, while end-users can share agent customizations with colleagues.
According to the Wall Street Journal, Nous generated $36 million in annualized revenue as of mid-September. The company expects that number to top $100 million by year’s end. Nous will use its funding to expand the adoption of Hermes’ business editions.