Nous Research, the open-source AI company, confirmed on October 7, 2026 that it has raised a $90 million Series B round, reportedly at a $1.5 billion valuation. The announcement, first reported by TechCrunch, pairs the funding with the announcement of Hermes for Businesses, a forthcoming enterprise product built around the company’s Hermes Agent. The round brings Nous Research’s total reported funding to $158 million. Nous has not yet given a launch date or pricing for Hermes for Businesses.
The timing matters. Enterprise buyers have spent 2026 asking a blunt question: can an AI agent actually run a multi-step business process without a human re-checking every output? Nous Research is betting that the answer is yes, and that the company willing to publish its model weights in the open will win more trust from procurement teams than one that keeps everything behind a locked API. That is a contrarian pitch in a market still dominated by closed, proprietary agent platforms from the largest labs.
Don't miss new tech stories on Google
Add Tech Insider once in the Google app and our stories appear in your news suggestions.
Nous Research Raises $90 Million at a $1.5 Billion Valuation
According to TechCrunch’s October 7 report, the Series B round was led by Robot Ventures, with participation from Nvidia, Union Square Ventures, Menlo Ventures, Samsung, and 1789 Capital. The $1.5 billion valuation represents a significant jump for a company that, by its own account, began as an informal open-source research collective rather than a venture-backed startup from day one. Nous Research’s own fundraising note, published on its website, confirmed the $90 million figure and the Hermes for Businesses launch without independently restating the valuation figure TechCrunch reported.
In its public statement, Nous Research framed the raise around a mission rather than a product roadmap. “Nous Research raised $90 million to create AI that serves the people,” the company wrote in its fundraise announcement. That language is notable for a company now taking money from Nvidia, Samsung, and a venture firm with direct ties to the Trump family, three investors whose interests do not always align neatly with a “for the people” framing of AI development.
| Detail | Figure |
|---|---|
| Round type | Series B |
| Amount raised | $90 million |
| Post-money valuation | $1.5 billion |
| Lead investor | Robot Ventures |
| Other investors | Nvidia, Union Square Ventures, Menlo Ventures, Samsung, 1789 Capital |
| Total funding to date | $158 million |
| Announcement date | October 7, 2026 |
| Flagship enterprise product | Hermes for Businesses |
Inside the Round: Who’s Backing Nous Research
Nvidia’s direct participation stands out. The chipmaker is both an investor in Nous Research and the dominant supplier of the GPUs that train and run its models, a dual role that has become common across the AI funding landscape in 2026 as compute providers take equity stakes in the companies buying their hardware. Samsung’s involvement points toward potential device or memory integration plays, an area the Korean conglomerate has been expanding into as on-device and edge AI inference demand grows.
The inclusion of 1789 Capital is the detail generating the most outside commentary. Donald Trump Jr. is a partner at the firm, and his participation in an AI funding round, even indirectly through the firm’s investment arm, has drawn attention given the broader political debate in Washington over AI regulation, export controls, and the Trump administration’s posture toward Silicon Valley. Nous Research has not commented specifically on the political optics of taking capital tied to the Trump family, focusing its public messaging instead on the technical and product story.
What Is Hermes Agent, and Why Does It Matter
Hermes Agent is Nous Research’s open-source system for handling multi-step tasks rather than single-turn chatbot exchanges. Where a conventional chatbot answers one prompt at a time, Hermes Agent is designed to break a request into a sequence of actions, carry state between steps, and adjust its approach based on intermediate results. Nous says it developed Hermes Agent during 2025 as an agent that can remember projects, create reusable skills, and carry knowledge between sessions, and it released the system under the MIT license in February 2026. The newest release is v0.22.4, a patch release for the CLI, installer, Microsoft Store app, Docker, and Hermes Cloud, published on October 21, 2026. On October 7, 2026, the company said Hermes had been cloned more than 24 million times and, by its own internal estimate, represented approximately 2.5% of global AI token usage, though these are company figures rather than independently verified ones.
That framing puts Hermes Agent squarely in the category industry watchers now call agentic AI: software that plans and executes rather than simply responds. The distinction matters for enterprise buyers because a single-turn chatbot carries limited operational risk, while an agent that can trigger actions across multiple systems carries real risk if it misinterprets a step. Nous Research’s bet is that an open system, where enterprises can inspect and modify the underlying weights, will be easier for security and compliance teams to trust than a closed agent whose internal reasoning is invisible.
The Hermes Model Family So Far
Hermes Agent did not appear out of nowhere. It builds on a model lineage that Nous Research has been iterating on publicly for more than two years. Hermes 3 launched in August 2024 as a set of full-parameter fine-tunes on top of Meta’s Llama 3.1, built to be highly steerable across instruction-following, reasoning, role-play, and tool use. Hermes 4 followed in August 2025 with a stronger emphasis on step-by-step reasoning. More recent reporting points to a Hermes 4.3 variant, reportedly a 36-billion-parameter model trained through Nous Research’s Psyche decentralized-training system, though that specific parameter count comes from a specialist secondary source rather than a primary Nous technical paper and should be treated as a reported figure rather than a confirmed spec.
| Release | Date | Notable detail |
|---|---|---|
| Hermes 3 | August 2024 | Full-parameter fine-tunes on Llama 3.1, open weights |
| Hermes 4 | August 2025 | Stronger step-by-step reasoning emphasis |
| Hermes 4.3 (reported) | 2026 | Reportedly ~36B parameters, trained via Psyche |
| Hermes Agent | 2026 | Multi-step agentic system, basis for enterprise product |
| Hermes for Businesses | October 7, 2026 | Enterprise deployment layer announced with Series B |
Hermes for Businesses: The New Enterprise Play
Hermes for Businesses is the product that justifies the enterprise framing of this round. Nous Research describes it as a way for organizations to deploy customized agents capable of handling multi-step workflows while keeping business data private and under the customer’s control, rather than routed through a third-party model provider’s infrastructure. “With our fundraising and new partners, we will build and deliver Hermes for Businesses,” the company said in its announcement.
That data-control pitch is not new in enterprise software, but it is a notable shift for a company whose identity has been built on open publication. Nous Research’s answer to that tension is a statement of philosophy rather than a technical specification: “We believe businesses of all sizes want the same liberties,” the company wrote, extending its open-source ethos from individual developers to corporate customers. Whether that framing survives contact with enterprise procurement, which typically demands SOC 2 reports, audit trails, and contractual liability terms rather than philosophical alignment, is an open question the company has not yet addressed in public materials.
From Open Collective to $1.5 Billion Startup: Nous Research’s Origin Story
Nous Research’s path to a $1.5 billion valuation did not start in a venture-backed incubator. The organization began in 2022 as an informal open-source research collective before incorporating as Nous Research, Inc. in May 2023. Reporting on the company identifies its founders as Jeffrey Quesnelle, Karan Malhotra, Ryan “Teknium,” and Shivani Mitra, a group that built its early reputation by publishing fine-tuned open models on Hugging Face and GitHub rather than pitching investors first. Current reporting identifies Dillon Rolnick as the company’s CEO.
That trajectory, open community to venture-backed unicorn in roughly three years, mirrors a pattern seen elsewhere in the open-weight AI race. It also explains why Nous Research’s public communications still read more like a researcher’s manifesto than a corporate press release, even as the company closes a nine-figure funding round with Nvidia and Samsung on the cap table.
DisTrO and Psyche: The Decentralized Training Bet Behind Hermes
Underpinning Nous Research’s technical identity are two projects: DisTrO and Psyche. DisTrO, short for Distributed Training Optimizer, is Nous Research’s approach to cutting the communication overhead that normally forces large model training onto tightly connected, high-bandwidth GPU clusters. By compressing the information GPUs need to exchange during training, DisTrO is designed to make it practical to train models across hardware that is not sitting in the same data center.
Psyche extends that idea into a full decentralized-training network, coordinating GPU capacity contributed from across the internet rather than requiring a single centralized cluster. Some reporting on Psyche describes it as using Solana-based coordination and incentive mechanisms to reward compute providers, though that blockchain-specific detail comes from secondary reporting rather than a primary Nous technical document and should be read as a reported characterization rather than a confirmed architectural fact. If Psyche works at scale, it gives Nous Research a structural cost advantage: access to distributed compute that does not require the kind of massive, centralized data center buildout that proprietary labs like OpenAI and Anthropic depend on.
The Trump Jr. Connection: 1789 Capital’s Role Draws Scrutiny
The presence of 1789 Capital on the investor list is unusual enough to warrant its own analysis. Donald Trump Jr.’s role as a partner at the firm means a Trump family member now has a financial stake in an open-source AI agent company at a moment when the federal government is actively shaping AI policy, from export controls on advanced chips to ongoing probes of major AI labs by agencies including the FTC. Nous Research’s fundraise note does not address the political dimension of the round directly, and the company has largely let the funding announcement speak in product and mission terms rather than political ones.
That silence is itself a data point. As AI companies increasingly find themselves at the intersection of national competitiveness narratives, especially around open-source models and China, investors with political visibility are becoming a more common, if quietly handled, feature of funding rounds across the sector.
Why Nvidia Is Investing Directly in an Open-Source Agent Startup
Nvidia’s participation gives the company a direct equity position in a firm built around open-weight agent models, while it simultaneously supplies the GPUs those models are trained and served on. That dual exposure, supplier and investor, has become one of the defining features of the 2026 AI funding market. For Nvidia, a bet on Nous Research is a bet on the open-weight segment of the agent market continuing to grow, which in turn supports continued GPU demand regardless of whether the long-term winner is a closed lab or an open one.
It also reinforces a trend covered in Nvidia’s broader $20 billion open-source push, where the company has been placing capital behind open-weight model developers as a hedge against the risk that China’s open-source labs come to dominate the category Nvidia’s hardware depends on for sustained demand.
Nous Research vs. the Enterprise AI Agent Field
Nous Research is not entering an empty market. Enterprise AI agents have become one of the most contested categories in software, with well-funded players like Sierra, Cognition, Decagon, Glean, and Harvey all pursuing variations of the same pitch: let an AI agent handle a complex, multi-step business process instead of a human. Reliable, independently verified 2025-2026 funding figures are not available for all of these companies, but Sierra’s position in the category has been widely referenced by AI funding trackers at a reported $950 million raised, a figure that should be treated as a secondary-source estimate pending confirmation from Sierra or a major financial outlet.
What differentiates Nous Research from most of that field is the open-weight commitment. Sierra, Cognition, Decagon, Glean, and Harvey all operate primarily as closed, proprietary platforms. Nous Research is trying to apply the open-source distribution model that has worked for its Hermes model releases to a category where enterprise buyers have historically preferred closed systems with clear vendor accountability. That is the central experiment of Hermes for Businesses: can an open agent framework win enterprise trust the way open-weight models have won developer trust.
| Company | Model/approach | Funding status |
|---|---|---|
| Nous Research | Open-weight Hermes Agent | $90M Series B, $1.5B valuation (Oct. 7, 2026) |
| Sierra | Closed conversational agent platform | Reported ~$950M raised (secondary-source estimate, unconfirmed by Sierra) |
| Cognition (Devin) | Closed autonomous coding agent | No independently verified 2025-2026 figure available |
| Decagon | Closed customer-support agent platform | No independently verified 2025-2026 figure available |
| Glean | Closed enterprise search and agent layer | No independently verified 2025-2026 figure available |
| Harvey | Closed legal-domain AI agent | No independently verified 2025-2026 figure available |
A Crowded Week for AI Funding: Armadin, Photon, and the October Rush
Nous Research’s round did not land in isolation. The same reporting window saw Armadin raise $255.5 million at a $2.5 billion valuation, and Photon close a $4.5 million seed round on October 1. Three distinct AI funding events inside a single week, spanning seed to late-stage rounds, is a signal of how much capital is still chasing the category even as public market observers debate whether AI valuations have outrun revenue. For context on the scale of capital flowing into open-weight competition specifically, Reflection AI’s reported $25 billion open-source bet targeting China’s lead in open models shows just how large the stakes have gotten at the top of the market, even as Nous Research operates at a comparatively modest, though still substantial, $1.5 billion scale.
The clustering of rounds also points to a broader pattern: investors are no longer betting on a single winner-take-all foundation model company. Instead, capital is spreading across a layered stack, foundation model labs, open-weight specialists, and now agent-deployment companies like Nous Research that sit on top of model weights rather than only producing them.
Historical Context: From Model Releases to Agent Platforms
Nous Research’s shift from publishing Hermes models to selling Hermes Agent as an enterprise platform mirrors a broader industry transition that accelerated through 2025 and into 2026. Early open-source AI labs built reputations by releasing fine-tuned models for the community to use and build on. The commercial challenge for all of them has been the same: open weights are easy to copy and hard to monetize directly. Wrapping those weights in an agent platform, with workflow tooling, data controls, and a subscription model, gives a company like Nous Research a path to recurring enterprise revenue that a raw model download never could.
That same transition has shaped moves elsewhere in the market, including Amazon’s open-source Strands Decider agent model, which similarly pairs an open release with a path toward production agent deployments rather than treating the model as the end product.
Market Impact: What This Means for Enterprise AI Buyers
For enterprise technology buyers, Nous Research’s entry adds a new option to an already crowded agent procurement decision. Companies evaluating AI agents for internal workflows now have to weigh open-weight flexibility against the operational maturity of closed platforms that have had more time to build compliance certifications, support organizations, and integration ecosystems. Hermes for Businesses is new enough that it has not yet published the kind of security documentation, SLAs, or case studies that enterprise buyers typically require before a production rollout.
That gap also creates risk for enterprises that move early. Deploying agents built on a model stack still being actively reshaped, Hermes Agent today, potentially a different architecture tomorrow as Psyche-trained models mature, means buyers need stronger internal testing and rollback plans than they would with a more established vendor. That operational risk is part of the reason security teams have started paying closer attention to how AI agent credentials and access get secured before agents are given production-level permissions inside a company’s systems.
The Open-Source vs. Proprietary Agent Debate
Nous Research’s enterprise push reopens a debate that has run through the entire AI industry since the first open-weight large language models appeared: does openness build more trust, or does it simply create more attack surface? Security researchers have increasingly flagged AI agent deployments as a distinct risk category, separate from traditional software, precisely because agents can take actions rather than only generate text. That concern has driven a wave of investment into AI agent security tooling, an area covered in detail around the funding gap between Zenity, HiddenLayer, and Straiker, three companies building tools specifically to monitor and constrain what AI agents are allowed to do once deployed.
Nous Research’s argument, implicit in its “liberties” framing, is that visibility into model weights and training methodology gives enterprises more ability to audit an agent’s behavior than they would get from a closed system they cannot inspect. Critics of that position argue that open weights also make it easier for bad actors to strip safety fine-tuning or repurpose an agent for unintended uses, a tension the broader open-weight AI industry has not resolved.
Expert and Industry Reactions
Nous Research itself has been the most direct voice on the record so far, with its public statements doing most of the work of explaining the strategic logic behind the round. The company confirmed the raise and its product direction directly on its social channels, posting that, “As reported in the @WSJ, we have raised a Series B to bring Hermes Agent to new frontiers (and, yes, build a mobile app),” according to a public post from Nous Research. That aside about a mobile app is a small but telling detail: it signals the company sees consumer-facing distribution, not just enterprise contracts, as part of how Hermes Agent grows from here.
Taken together, Nous Research’s own statements paint a company trying to hold two identities at once, open-source research collective and venture-backed enterprise vendor, without fully choosing between them. “We believe businesses of all sizes want the same liberties,” as the company put it, is as much a statement of brand positioning as it is a product claim, and it will be tested the moment Hermes for Businesses faces its first enterprise security review.
What Happens Next: 5 Predictions for Nous Research and Enterprise Agents
Several developments look likely in the months following this announcement, based on how similar rounds and product launches have played out elsewhere in the AI agent market this year.
- Enterprise security documentation becomes the next test. Hermes for Businesses will need published compliance and audit materials before large enterprises move past pilot deployments, and that process typically takes several quarters even for well-funded startups.
- More open-weight labs will chase the agent-platform pivot. Nous Research is unlikely to be the last open-source model developer to wrap its weights in a paid enterprise agent layer rather than relying on model downloads alone for revenue.
- Scrutiny of politically connected investors will grow. As AI policy debates intensify in Washington, expect more attention on investors like 1789 Capital whenever a fundraising round intersects with figures close to the current administration.
- Nvidia’s investor role in AI startups keeps expanding. Direct equity stakes alongside hardware supply relationships are likely to become a standard feature of late-stage AI rounds through the rest of 2026.
- Competitive pressure on AI agent security tooling will rise. As more open and closed agent platforms reach enterprise production, demand for independent monitoring and access-control tools built specifically for agents should keep climbing.
None of these are guarantees, and Nous Research’s own public materials have not laid out a detailed roadmap beyond the Hermes for Businesses launch itself. But the pattern across the broader AI funding market in October 2026, three sizable rounds inside a single week, a chipmaker investing directly in model developers, and an accelerating race to convert open models into paid agent products, makes these the most likely near-term developments to watch.
Frequently Asked Questions
How much did Nous Research raise, and at what valuation?
Nous Research raised $90 million in a Series B round announced on October 7, 2026, valuing the company at $1.5 billion, according to TechCrunch. The round brings the company’s total funding to $158 million.
Who led Nous Research’s Series B round?
Robot Ventures led the round, with participation from Nvidia, Union Square Ventures, Menlo Ventures, Samsung, and 1789 Capital.
What is Hermes Agent?
Hermes Agent is Nous Research’s open-source AI system designed to carry out multi-step tasks rather than respond to single prompts, positioning it as an agentic alternative to conventional chatbots.
What is Hermes for Businesses?
Hermes for Businesses is the enterprise product Nous Research announced alongside its Series B, designed to let organizations deploy customized AI agents for multi-step workflows while keeping their data private and under their own control.
Why is Donald Trump Jr.’s involvement significant?
Donald Trump Jr. is a partner at 1789 Capital, one of the investors in Nous Research’s Series B. His indirect participation has drawn attention given the ongoing political debate over AI regulation and the Trump administration’s relationship with the AI industry.
How old is Nous Research, and who founded it?
Nous Research began as an informal open-source research collective in 2022 and incorporated as Nous Research, Inc. in May 2023. Reporting identifies its founders as Jeffrey Quesnelle, Karan Malhotra, Ryan “Teknium,” and Shivani Mitra.
What are DisTrO and Psyche?
DisTrO, or Distributed Training Optimizer, is Nous Research’s method for reducing the data-transfer burden of training large models across distributed hardware. Psyche extends that approach into a full decentralized-training network that coordinates GPU capacity contributed across the internet, with some reports describing Solana-based coordination for participating compute providers.
How does Nous Research compare to closed AI agent startups like Sierra or Cognition?
Unlike Sierra, Cognition, Decagon, Glean, and Harvey, which operate closed, proprietary agent platforms, Nous Research is trying to apply its open-weight model philosophy to the enterprise agent category, betting that transparency around model weights will build enterprise trust in a way closed systems cannot.
