Nous Research Raises $90M to Take Its Open-Source Hermes Agent to Business

Nous Research, the company behind the open-source Hermes AI agent, has raised $90 million. Chief executive Dillon Rolnick announced the raise in a note on the company’s website, saying the money will build “Hermes for Businesses”, a version of the agent for companies [1].

The raise is confirmed by Nous [1]. The valuation is not: Nous did not state one. Press reports mostly put it at $1.5 billion, though a $1.2 billion figure has also circulated [2][3][4][6]. Nous’s claim that Hermes drives about 2.5% of global AI token usage is the company’s own estimate [1].

What has Nous announced?

In its note, Nous says it raised $90 million and names investors including NVIDIA, Microsoft’s M12 venture fund, Samsung, Robot Ventures, Union Square Ventures, Y Combinator and Menlo Ventures, “among others” [1].

TechCrunch reports the round is a Series B led by Robot Ventures, with participation that also includes 1789 Capital, where Donald Trump Jr. is a partner. It says the round brings the three-year-old company’s total funding to $158 million [2].

Rolnick’s note is partly a manifesto. Nous was formed in 2023 to push back against AI becoming centralised and gate-kept, he writes, and he argues the new capital will build an alternative that is “open, competitive, ‘safer’ because of its auditability and not because of its gate-keepers” [1]. He adds: “Somehow, $90 million is modest in modern fundraising” [1].

What is the valuation?

This is where the labels matter.

• $1.5 billion, reported. The Wall Street Journal, which first reported the round, says Nous raised $90 million “at a $1.5 billion valuation” [3]. TechCrunch also reports $1.5 billion and says the announcement confirms its own earlier reporting [2]. SiliconANGLE reports the same figure [4].
• $1.2 billion, also in circulation. News aggregator Exact.news summarised the Journal’s report as a $1.2 billion valuation [6]. The Journal article as we read it on 8 October gives $1.5 billion [3], so the lower figure may reflect an early or mistaken summary. We could not confirm it from any primary source.
• Nous’s own figure: none. The company’s note does not mention a valuation [1], and The Next Web noted that Nous did not give one [5].

The fair summary: $90 million raised (confirmed), at a valuation reported by the Journal and TechCrunch as $1.5 billion, with one $1.2 billion figure in circulation.

What is Hermes Agent?

Hermes Agent is an AI agent: software that uses AI models to carry out tasks, not just answer questions. Nous says it built Hermes to learn as it goes. It remembers projects, writes reusable “skills” as it works, and carries that knowledge across sessions. Users choose where it runs and which AI models it uses [1].

Nous released Hermes under the MIT licence, a permissive open-source licence, in February 2026 [1]. SiliconANGLE describes it as a desktop app that can work with a user’s files and programs, use a browser, and run code in a container, with a router that sends each prompt to the model best suited to it, weighing cost, speed and quality [4]. It can use popular commercial models such as Claude as well as Nous’s own fine-tuned open models [4].

How popular is it?

Nous makes two adoption claims in its note [1]:

• Hermes “has been cloned over 24 million times”.
• “To our best internal estimates,” it “drives approximately 2.5% of all global token usage”. A token is a small chunk of text processed by an AI model, so this is a claim about Hermes’s share of all AI processing worldwide.

Both are Nous’s claims. We have seen no independent measurement of the 2.5% figure, and it is not obvious how anyone outside could measure global token usage precisely. The Wall Street Journal reported a different adoption metric: more than 22 million downloads since February [3].

The Journal also reported that Nous was at about $36 million in annualised revenue by mid-September and expects to pass $100 million before the end of 2026, as relayed by TechCrunch and SiliconANGLE [2][4]. Those are reported figures; we could not read the full Journal article to see how it sourced them.

What is Hermes for Businesses?

Nous says the new product will let businesses choose “the best models at the best prices” for their goals, scale AI use “safely and privately”, and build up and own their institutional knowledge [1]. TechCrunch describes it as letting companies deploy customised agents that handle multi-step workflows while keeping data private and secure [2]. SiliconANGLE reports existing paid editions, Hermes Cloud, Hermes Business and Hermes Enterprise, with team-level controls on token use [4].

One related detail from this week: Microsoft lists “Hermes Agent by Nous Research” among the agents that will be releasing support for its new Windows agent sandbox, Microsoft Execution Containers [7]. For a product pitched at companies, being able to run inside an operating-system-enforced boundary is likely to matter.

Update, 8 October 2026

Nous has also launched a leaderboard. On 6 October it published the Hermes Index, which it calls “an opinionated index of agentic capability” [8]. Nous says the goal is “to better inform users choosing between models and to show labs how they perform in Hermes” [9].

The index runs each model inside Hermes Agent on four test suites: Nous’s own Hermes Bench (150 tasks across 25 categories), TerminalBench 4, TerminalBench Science and SkillsBench. It reports the average score and the average cost per task for 14 models [8]. The top three, all Nous’s own figures [8]:

• Claude Opus 5.5: 63.31, at $4.99 per task (cost marked provisional).
• GPT-6 Astra: 56.25, at $11.61 per task.
• Claude Sonnet 5.5: 53.14, at $2.82 per task (cost marked provisional).

At the cheap end, Ling 3.0 Flash scores 21.56 at $0.054 per task [8].

Cost per task is a useful extra view for anyone choosing a model, since the most expensive model here is not the top scorer. But this is not a neutral benchmark. Nous built the software the tests run in, wrote one of the four suites and sells the agent [8]. Its own notes add that some SkillsBench runs found that suite’s public answers online; Nous drops those tasks from the score that counts [8]. Treat the index as a guide to how models perform inside Hermes, not a general ranking.

Why it matters

Personal AI agents have drawn large sums this year, as The Next Web notes [5]. Nous is betting that companies will prefer an agent whose code they can inspect, run where they choose, and pair with any model. That is the same argument open-source software has made for decades, now applied to agents.

The test is whether a project popular with developers and hobbyists can turn into something IT departments trust and pay for.

What this does not prove

• The valuation. Nous did not state one. $1.5 billion is reported by the Journal and TechCrunch; $1.2 billion appeared in an aggregator summary [2][3][6].
• The 2.5% token-usage share. That is Nous’s internal estimate [1].
• That clones equal users. “Cloned 24 million times” (Nous) and “22 million downloads” (WSJ) are not the same as active users [1][3].
• The revenue figures. They are the Journal’s reporting, which we could only see second-hand [2][4].
• That Hermes for Businesses is ready. Nous describes what it will build; we have not seen it in use.
• That the Hermes Index is an independent ranking. It is run by Nous, inside Nous’s own agent, partly on Nous’s own test suite [8].

The Bottom Line

Nous Research has confirmed a $90 million raise to turn its MIT-licensed Hermes agent into a business product [1]. The valuation is reported, not company-stated: $1.5 billion according to the Wall Street Journal and TechCrunch, with a $1.2 billion figure seen in one aggregator summary [2][3][6]. The striking claim that Hermes drives about 2.5% of global AI token usage is Nous’s own estimate [1]. The open-agent bet has serious backing; what it needs next is paying business customers.

Sources

1. Dillon Rolnick, “A Note on Our Fundraise,” Nous Research, October 2026. https://nousresearch.com/a-note-on-our-fundraise
2. Marina Temkin, “Nous Research confirms it hit $1.5B valuation, launches AI agents for business users,” TechCrunch, 7 October 2026. https://techcrunch.com/2026/10/07/nous-research-confirms-it-hit-1-5b-valuation-launches-ai-agents-for-business-users/
3. Yuliya Chernova, “Nous Research Scores $90 Million to Bring Open-Source AI Assistant to Enterprises,” The Wall Street Journal (WSJ Pro), 7 October 2026 (paywalled; opening paragraphs read). https://www.wsj.com/pro/venture-capital/nous-research-scores-90-million-to-bring-open-source-ai-assistant-to-enterprises-516f9208
4. Maria Deutscher, “Nvidia, Samsung back $90M round for AI agent startup Nous Research,” SiliconANGLE, 7 October 2026. https://siliconangle.com/2026/10/07/nvidia-samsung-back-90m-round-for-ai-agent-startup-nous-research/
5. Cristian Dina, “Nous Research raises $90M to take its Hermes AI agent to businesses,” The Next Web, 7 October 2026. https://thenextweb.com/news/nous-research-90m-1-5bn-hermes-agent-wsj
6. Exact.news, “Nous Research raises $90 million to expand its open-source Hermes agent into enterprises” (aggregator summary citing a $1.2 billion valuation), 7 October 2026. https://exact.news/story/nous-research-raises-90-million-for-open-source-hermes-agent
7. Logan Iyer, “Microsoft Execution Containers: Policy-driven containment for AI agents,” Windows Developer Blog, 7 October 2026. https://blogs.windows.com/windowsdeveloper/2026/10/07/microsoft-execution-containers-policy-driven-containment-for-ai-agents/
8. Nous Research, “Hermes Index | Agentic Model Leaderboard,” Nous Research portal, read 8 October 2026. https://portal.nousresearch.com/bench
9. Nous Research (@NousResearch), “Introducing Hermes Index…,” post on X, 6 October 2026, 23:02 BST. https://x.com/NousResearch/status/2107592255124951360