Hi - I answer from the OpenSmartRoute documentation: routing, the API, plans and quotas, self-hosting. Ask away, or open a support ticket if you need a person.
Grounded in the docs - follow a source before acting on it.
Nvidia and Samsung back Nous Research with $90 million funding - OpenSmartRoute
Nvidia Corp and Samsung Electronics joined Robot Ventures to back Nous Research. The startup raised $90 million in a Series B funding round. This money supports the development of its Hermes artificial intelligence agent. Nous Research Inc. is now valued at $1.5 billion after this announcement. Early-stage startup fund Robot Ventures led the investment round. Y Combinator and several other investors also joined the round.
This funding signals strong confidence in the company's technology. Nvidia Corp has been a major player in AI hardware for years. Samsung Electronics Co. is a global leader in semiconductors and electronics. Their support suggests they see value in Nous' agent platform. Robot Ventures focuses on early-stage startups with high growth potential. The total funding amount of $90 million is significant for an early-stage company.
Nous Research Inc. was founded to build advanced AI agents. They focus on creating tools that interact with user files and data. The Hermes agent allows users to organize documents or run code automatically. Before this round, the company relied on its open-source license model. Now, they have capital to expand their business editions for organizations.
The news comes from SiliconANGLE AI on October 7, 2026. This date marks a key moment in the AI agent startup landscape. Many companies are racing to build agents that can perform complex tasks. Nous Research aims to lead this race with its Hermes platform. The funding will help them scale their adoption of business editions.
The company released Hermes in February under an open-source license. Since then, the agent has been downloaded more than 24 million times. It boasts several thousand open-source contributors who have helped improve the software. This number of downloads shows immense popularity among developers and users worldwide.
Token consumption is a critical metric tied to AI models' hardware use. Hermes accounts for an estimated 2.5% of the world's token consumption. This percentage represents the total amount of data processed by the model. High usage indicates that many people rely on this agent for daily tasks.
South Korea plans to spend $3.49 billion on a domestic frontier AI model. The government also wants to encourage companies to use these models across the economy.
Mirror Particle raises capital to create an AI engine that simulates changing human motivations. The company rejects large language models in favor of a foundation model trained on longitudinal data.
Open-source licenses allow anyone to download and modify the software freely. Developers can integrate Hermes into their own projects without paying licensing fees. This openness has driven rapid adoption across different industries. The community contributions ensure the software stays updated with new features.
Token consumption measures how much data an AI model processes during operation. It directly correlates with the energy and hardware resources required to run the model. A 2.5% share is a massive portion of global AI usage. This highlights the efficiency and demand for Hermes' specific capabilities.
Users can verify these metrics through public dashboards or documentation. The team likely publishes reports on token usage trends over time. Monitoring this data helps developers understand system load patterns. Engineers can use this information to plan their own infrastructure needs.
The combination of high downloads and significant token usage proves the agent's utility. It serves as a benchmark for other AI agents in terms of adoption. Developers often compare their projects against popular open-source alternatives. Hermes sets a high standard for what an agent platform can achieve.
Hermes capabilities - desktop app features, tool use, and model integration
Hermes ships as a desktop app that can interact with the user's files. A consumer could ask the agent to organize a collection of documents by category. The software accesses programs running on the user's local machine directly. It uses a browser to find web data needed for specific tasks.
The agent can run code inside a Docker container for secure execution. This capability allows Hermes to perform complex operations without risking system stability. Users benefit from automation that handles repetitive file management tasks efficiently. The tool use extends beyond simple text generation into actual system interaction.
Under the hood, Hermes uses popular large language models such as Claude. It completes tasks for users by leveraging the reasoning power of these models. The system also supports a lineup of LLMs developed specifically by Nous. These are fine-tuned or customized versions of open-source LLMs.
Large language models (LLM) are advanced AI systems capable of generating human-like text. They process input prompts and produce relevant outputs based on training data. Fine-tuning adjusts the base model to perform better on specific tasks or domains. Customization allows developers to tailor the model's behavior to unique needs.
Hermes integrates multiple models to handle diverse user requests effectively. It switches between Claude and Nous models depending on the task complexity. This flexibility ensures high performance across different types of interactions. The desktop app provides a unified interface for all these capabilities.
Users can test these features by installing the Hermes desktop application. Local execution means data stays on their machine unless they choose otherwise. Security is enhanced because sensitive files do not leave the device unnecessarily. Developers can build custom workflows around these built-in tools and integrations.
Model development - Hermes 4 series, DataForge, and synthetic dataset creation
Nous' latest collection of large language models is called the Hermes 4 series. The company customized these models using a training dataset with 5 million synthetically-generated records. About two thirds of these records were specifically created to enhance reasoning capabilities.
The custom tool called DataForge generated most of this specialized training data. Synthetic data creation allows companies to generate vast amounts of high-quality examples. This approach helps overcome the limitations of relying solely on public datasets. Reasoning capabilities are crucial for agents that need to solve complex problems.
The Hermes 4 series represents a significant step forward in model performance. It builds upon previous versions by incorporating more advanced synthetic data techniques. Fine-tuning with this data improves the models' ability to follow instructions precisely. Users expect better accuracy when interacting with newer model generations.
Synthetic datasets can simulate rare or edge-case scenarios that real-world data lacks. This ensures the model is prepared for unexpected inputs from users. DataForge automates the process of creating these challenging training examples efficiently. The result is a more robust and reliable AI system overall.
Developers can evaluate the improvements in the Hermes 4 series against older versions. Benchmarks often show measurable gains in logic, math, or coding tasks. The synthetic data approach is becoming a standard practice in model development. It reduces the time needed to train models from scratch significantly.
Router and personalization - how prompts are routed and user preferences handled
A so-called router enables Hermes to send each user prompt to the best-equipped LLM. This mechanism takes into account factors such as inference costs, speed and output quality. The goal is to deliver the fastest and most accurate response possible for every request.
Inference cost refers to the money spent running a model on hardware. Speed measures how quickly the model generates its output response. Output quality determines how well the answer matches what the user intended. Balancing these three factors optimizes the overall experience for the end-user.
Another built-in module personalizes LLM responses based on user preferences stored locally. Optionally, users can enhance these features by connecting to external databases like RetainDB. The agent-optimized RetainDB relational store helps manage structured data efficiently.
RetainDB is a specialized database designed for AI agents to store and retrieve information quickly. Adding it enables Hermes to retain preferences and project data across user sessions. This continuity makes the agent feel more familiar and helpful over time.
User preferences might include preferred tone, specific formatting rules, or historical context. The personalization module learns from past interactions to adapt future responses accordingly. This creates a smoother workflow as the agent anticipates user needs better.
Engineers can configure routing rules to prioritize certain models for specific tasks. They set weights for cost, speed, and quality in the router configuration. Hard rules ensure sensitive data never leaves an on-premises model environment. Such controls are essential for maintaining data privacy and security standards.
Monetization and revenue - paid editions, annualized revenue, and future projections
The company monetizes its agent with three distinct paid editions available for purchase. The first edition is Hermes Cloud, which lets consumers launch agents in seconds. Users do not need to set up any infrastructure when using this service.
Hermes Business and Hermes Enterprise are geared towards organizations needing advanced features. Administrators can set team-level token use limits within these services. End-users can share agent customizations with colleagues for collaborative work environments. This tiered approach allows different groups to pay based on their specific needs.
According to the Wall Street Journal, Nous generated $36 million in annualized revenue as of mid-September. The company expects this number to top $100 million by year's end. This growth trajectory demonstrates strong market demand for their paid services. Revenue comes from subscriptions and enterprise licensing agreements primarily.
Annualized revenue smooths out monthly fluctuations to show consistent yearly performance trends. It helps investors and partners understand the long-term financial health of the company. Reaching $100 million annually would place Nous among top AI startups financially. The funding round will directly support expanding these business editions further.
Funding allows the team to invest in sales teams and customer support infrastructure. They can offer better training for enterprise clients adopting Hermes Enterprise. Marketing efforts will focus on highlighting the efficiency gains from automated agents. Partnerships with cloud providers might also accelerate adoption among large organizations.
Managers should look at these revenue figures when evaluating potential AI agent investments. A company projecting $100 million in revenue shows strong business viability. The shift from open-source to paid models indicates a mature product ready for scale. Investors often prefer companies with clear monetization strategies over pure research labs.
Why it matters - impact on cost, speed, quality, and new capabilities for users
This news matters because it validates the commercial potential of advanced AI agents. Cost savings become possible when routers select cheaper models for simple tasks. Speed improves as the system avoids using heavy models unnecessarily. Quality rises when the right model handles complex reasoning problems.
New capabilities emerge as companies adopt Hermes' router and personalization features. Organizations can automate workflows that were previously too expensive or slow to build manually. The ability to route requests dynamically reduces operational overhead significantly. Users gain access to powerful tools without needing deep technical expertise.
For engineers, this means more efficient ways to manage AI infrastructure at scale. Managers see a clear path to reducing costs while improving productivity metrics. Safety and compliance features like on-premises routing protect sensitive data better. The combination of open-source flexibility and paid enterprise options offers unique value.
The impact extends beyond Nous Research to the broader AI ecosystem. Competitors will likely adopt similar routing and monetization strategies soon. Standard practices for managing LLM costs are becoming more defined globally. This trend encourages innovation in how we deploy and utilize large language models.
Users who rely on agents for daily tasks benefit from increased reliability and speed. Businesses that automate routine work see immediate returns on their technology investments. The availability of multiple model tiers ensures everyone finds a suitable solution. It lowers the barrier to entry for adopting AI automation tools widely.
How OpenSmartRoute helps - routing decisions based on inference costs and speed
A team routing requests through OpenSmartRoute gains precise control over model selection. They can score every candidate on quality, cost, speed and safety metrics. The platform allows setting weights per request to prioritize specific factors dynamically. Hard rules pin a request to ensure compliance with internal policies.
Text with personal data stays on an on-premises model within the OpenSmartRoute setup. A region or a cost cap is never crossed without explicit configuration changes. This prevents accidental data leaks or budget overruns during high-traffic periods. The system learns from outcomes so successful models get more traffic automatically.
OpenSmartRoute works with any OpenAI-compatible provider alongside open-weight models served locally. It supports MCP tools and A2A agents for extended functionality beyond standard chat. A savings ledger shows what each routed request cost next to the most expensive option. This transparency helps teams justify infrastructure spending to stakeholders clearly.
The input guard spots prompt injection and personal data before a request leaves the environment. 'osr eval' measures routing accuracy on the team's own prompts and can fail a build when it drops. These features ensure the router remains secure and accurate over time. The hosted platform keeps a models catalogue with prices and public rankings built from real traffic.
Teams using OpenSmartRoute see immediate improvements in operational efficiency and cost management. They avoid paying for capabilities they do not need for specific tasks. The ability to customize routing rules fits diverse organizational requirements perfectly. It scales as the number of users and models grows significantly.
What to do - actions engineers and managers can take with this news
Engineers should evaluate their current AI infrastructure against Hermes' router architecture. They might consider implementing a similar dynamic routing system for their own projects. Comparing token consumption metrics helps identify opportunities for cost optimization quickly. Testing synthetic data generation could accelerate their model development cycles too.
Managers should review the revenue projections and funding rounds when planning budgets. Investing in agent platforms with clear monetization paths often yields better ROI. Teams can set up pilot programs to test Hermes' business editions before committing fully. Understanding the $36 million to $100 million revenue range sets realistic expectations.
Both groups should investigate open-source alternatives for their immediate needs. OpenSmartRoute offers a practical way to start without heavy upfront costs. They can use the router to experiment with different models safely. The savings ledger provides concrete data to present to leadership teams later.
Monitoring token consumption is another actionable step everyone can take immediately. Tools exist to track usage patterns and identify expensive model deployments. Reducing unnecessary inference saves money and lowers environmental impact significantly. Engineers can set up alerts for anomalous spending behavior proactively.
Finally, engaging with the community around Hermes and similar projects fosters collaboration. Sharing insights about routing strategies helps improve industry standards collectively. Contributing to open-source initiatives strengthens the ecosystem for everyone involved. Staying informed on funding rounds keeps teams ahead of emerging trends.
