AgentsNews

Nous Research raises $90M and launches Hermes for Businesses, an enterprise version of its open-source agent

Nous Research announced a $90 million Series B on October 7, 2026, at a $1.5 billion valuation according to TechCrunch, and is turning its MIT-licensed Hermes Agent into a paid team product. What Hermes for Businesses includes, and what the "cloned 24 million times" claim does and doesn't mean.

A glass bar and a taller bar with a lit top section at true ratio on a fuchsia gradient, titled $90 million

What’s new, in 20 seconds

  1. On October 7, 2026, Nous Research CEO Dillon Rolnick said the company raised $90 million. TechCrunch reported it is a Series B led by Robot Ventures at a $1.5 billion valuation, bringing total funding to $158 million. Nous itself didn't state a valuation.
  2. The money goes to Hermes for Businesses: a self-serve Business tier on Nous infrastructure, available now, and an Enterprise tier on customer infrastructure with SSO and SLAs, by inquiry. Nous hasn't published prices.
  3. Nous says Hermes "has been cloned over 24 million times" and estimates it drives about 2.5% of global token usage. Both are the company's own figures. It doesn't define "cloned", and the WSJ separately reported a download count, seen by 9to5AI only through secondary summaries (about 22.7 million).
Contents

On October 7, 2026, Nous Research said it had raised $90 million to build Hermes for Businesses, a paid team version of Hermes Agent, its free, MIT-licensed AI agent. TechCrunch reported the round is a Series B at a $1.5 billion valuation, confirming earlier reporting of the talks. The deal puts one of the most popular open-model agent projects onto the same path as other open-source companies: keep the core free and sell hosting, administration and support to companies.

This story is based on Nous Research's funding note and Hermes for Business product page, and on TechCrunch's and other outlets' reporting, checked October 12, 2026.

The short version

  • Funding: $90 million, announced by CEO Dillon Rolnick on October 7. TechCrunch reports a Series B led by Robot Ventures at a $1.5 billion valuation, with $158 million raised in total.
  • Product: Hermes for Businesses. A self-serve Business tier runs on Nous's infrastructure and is available now. An Enterprise tier runs on the customer's own infrastructure and is by inquiry.
  • Price: not published.
  • The "24 million" claim: Nous says Hermes "has been cloned over 24 million times." That's the company's number, and it doesn't say how it counts.

What did Nous Research announce?

Rolnick's note on the fundraise says: "Nous Research raised $90 million to create AI that serves the people." It names NVIDIA, Microsoft's M12, Samsung, Robot Ventures, Union Square Ventures, Y Combinator and Menlo Ventures "among others." The note doesn't give a valuation or say what the round is called.

TechCrunch filled in the rest on October 7:

  • it's a Series B led by Robot Ventures, at a $1.5 billion valuation
  • 1789 Capital also took part
  • the round brings total funding to $158 million
  • citing the Wall Street Journal, Nous had about $36 million in annualized revenue by mid-September and expects to pass $100 million before the end of 2026

The note is mostly a mission statement. Rolnick writes that "the conversation is being dominated by fearmongering pleas for sweeping regulation," and argues for AI that is "open, competitive, 'safer' because of its auditability and not because of its gate-keepers."

What is Hermes for Businesses?

The note promises that Nous "will build and deliver Hermes for Businesses," which will let companies choose models by price and goal, keep their AI use private, and "own their institutional knowledge." Nous's product page, titled "Hermes for Business," is more concrete. It offers two tiers.

BusinessEnterprise
Runs onNous infrastructure (an isolated tenant per team)Customer-controlled infrastructure
SetupSelf-serveTailored deployment, white-glove onboarding
ExtrasAdmin roles, per-member spend caps, API keysSSO, SLAs
AvailabilityAvailable nowRegister interest / contact Nous
PriceNot publishedNot published

Last verified: October 12, 2026 (Nous Research, Hermes for Business page)

Both tiers include shared "skills" that any team member discovers and that then become available to the whole organization. A single harness works across every model on Nous's inference API or with the team's own provider keys, and there's spending visibility from one central balance. The agent reaches staff through Discord, Telegram, Slack and WhatsApp, as well as the terminal, desktop apps and the cloud. Running one agent harness across several models is a form of model routing, with the cost trade-offs that come with it.

What is Hermes Agent?

Hermes Agent is Nous's open-source, self-improving agent ("The agent that grows with you"). Its GitHub repository is MIT-licensed and showed about 250,000 stars and 53,000 forks on October 12. Tech Funding News reports it was released in February 2026. You can self-host it with one-line installers for Linux, macOS, WSL2 and Windows. On October 8, Nous said it was also available in the Microsoft Store. It is a general-purpose agent, not just a coding tool. Our explainer on what "agentic" really means covers how this kind of tool loop works.

Has Hermes really been "cloned 24 million times"?

That's what Nous says, and only Nous. Rolnick's note reads: "Today, Hermes has been cloned over 24 million times." It adds that Hermes "drives approximately 2.5% of all global token usage," based on "our best internal estimates." TechCrunch and others repeated both figures as the startup's estimates.

The note doesn't define "cloned." It could mean git clone operations, installs or something broader. GitHub shows clone traffic only to repository owners, and only for the last 14 days, so the figure can't be checked from outside. The Wall Street Journal reported a different metric: Dealroom and Digg summarize it as about 22.7 million (or "more than 22 million") downloads since February. Read both as adoption signals reported by the company, not audited counts. The 2.5% token-share estimate has no published method behind it.

Why it matters

Hermes for Businesses tests whether companies will pay for a hosted, managed version of an agent they could run free. Tech Funding News put it directly: "The harder test is whether firms that can run Hermes for free will pay Nous to run it for them." The pitch is administration (spend caps, SSO, shared skills) and data control, not a better model. Enterprise buyers should also weigh the security side of always-on agents that read chat messages and documents. Our piece on the case against unattended agents covers how sandboxing limits that risk.

What's confirmed, reported and uncertain

  • Confirmed (Nous): the $90 million raise, the named investors, the Business and Enterprise tiers and their features, Business availability, and the MIT license.
  • Reported (TechCrunch, WSJ): Series B label, $1.5 billion valuation, Robot Ventures as lead, 1789 Capital, $158 million total, and roughly $36 million annualized revenue.
  • Company claims, unverified: "cloned over 24 million times" and about 2.5% of global token usage.
  • Unknown: pricing for either tier, an Enterprise launch date, and the planned mobile app's timing.

Read next

Comments (0)

More on Open models

View all
Open source · NewsNemotron 3.5 Transcribe: Nvidia's open speech recognition model covers 40 languages under OpenMDW, with about 4B active parametersNemotron 3.5 Transcribe, released by Nvidia on Hugging Face on October 28, 2026, is an open-weight speech recognition model for 40 languages under the OpenMDW-1.1 license. It activates about 4B parameters per token, runs on a single Nvidia GPU and handles recorded audio, not live streams. Nvidia's word error rates are vendor-reported.·6 min read
Research · AnalysisQwenputer: a developer compiled a RISC-V CPU into Qwen3.5-4B's weights, with no training, and says it runs 18 million instructions without an errorQwenputer is Qwen3.5-4B with a 32-bit RISC-V CPU compiled directly into 11 extra layers, using the model's Gated DeltaNet memory as RAM. Developer Divyansh Agrawal says no weights were trained, the CPU passes the official RISC-V tests, and Qwen can call it as a tool for exact arithmetic. What was built, how, what 'deterministic' means and the limits.·7 min read
Open source · ExplainerWhat is an abliterated model? How refusal removal works, why abliterated DeepSeek is now sold by API, and the safety questionsWhat is an abliterated model? It's an open-weight model whose weights have been edited to strip out its learned refusals, based on 2024 research showing refusal sits along a single direction inside the model. In October 2026, an API router began selling an abliterated DeepSeek V4.1 Flash from five providers. What the technique is, what the licenses and terms allow, and what safety researchers say.·8 min read
Products · ExplainerOffline AI assistant: how Survival LLM pairs a local model with a 930,000-article emergency library, and how people run AI with no internetAn offline AI assistant runs the model and its reference library on your own computer, so it keeps working when the internet doesn't. Survival LLM, which drew wide attention on October 27, 2026, bundles one of four models with about 930,000 Wikipedia articles and medical and repair references. What it needs, how Kiwix, Ollama, llama.cpp and LM Studio work offline, and where offline AI falls short.·9 min read
The Week in AI · every Friday

Get the cluster, not just the headline.

The week’s biggest stories in AI and tech, what they mean, and what to read next.

Free. One email a week. Unsubscribe in one click.