Nous Research
Open-source LLMs and decentralized AI research collective
Performances
RD Popularity Index
RD Growth Index
Comparison
Follow Updates
Follow Lists
People
About Nous Research
Nous Research is a decentralized AI acceleration company focused on open-source models and distributed training infrastructure.
Its core business includes the Hermes series of open-source LLMs, the Hermes Agent framework for multi-platform AI assistants, and the Psyche network for decentralized training over the internet. The company's key differentiator is its DisTrO optimizer, which drastically reduces data transfer during distributed training, making collaborative GPU training across regions practical. Hermes models are also known for a less restrictive design philosophy, with a non-refusal rate around 60% versus roughly 5% for mainstream open models.
The market pain point addressed is the centralization of AI compute and excessive model censorship. Nous aims to prove that decentralized training is a viable alternative to concentrated data center buildouts, while offering models with fewer arbitrary refusals.
In the past six months (March to September 2026), Hermes Agent added browser automation via a single unified tool (replacing a dozen schemas), cutting token usage by 48-66% with no accuracy loss. Desktop plugins arrived (Kanban was the first official one), voice activation was added, and a credential firewall for Docker sandboxes was introduced. The Nous Portal cloud hosting service also launched, allowing users to spin up a working agent in under a minute, with subscription tiers from $20 to $200 per month.
One notable negative: tokens named "NOUS" trading on Solana DEXs are not official issuances from Nous Research, posing a scam risk to investors.
Founded in 2023 by Jeffrey Quesnelle, Karan Malhotra, and Ryan Teknium, Nous Research emerged from an open-source AI research community. Its core mission is advancing open-source language models and decentralized AI infrastructure.
In December 2024, the team publicly live-streamed the distributed pre-training of a 15B-parameter language model using its self-developed DisTrO optimizer, demonstrating the feasibility of decentralized training across geographic regions. This marked a technical milestone.
In April 2025, Nous completed a $50 million Series A led by Paradigm at a $1 billion token valuation, with earlier seed funding of $20 million from Distributed Global and OSS Capital. The capital supported development of Psyche Network, a Solana-based decentralized training system.
By mid-2026, the company was in talks to raise $75 million at a $1.5 billion valuation, led by Robot Ventures with USV participating. Its Hermes model achieved 214,000 GitHub stars, nearly 40,000 forks, and 271 billion daily token consumption on OpenRouter, ranking first among AI agents. These milestones reflect Nous's trajectory from research community to a leading force in open-source, decentralized AI.
Over the past six months (2026-03-01 to 2026-09-01), Nous Research has made significant strides in decentralized AI. In July 2026, the company closed a new funding round led by Robot Ventures, raising at least $75 million at a valuation of approximately $1.5 billion, following its earlier $50 million Series A led by Paradigm.
Product-wise, Hermes Agent has evolved into a full-fledged autonomous AI agent platform, offering both local and cloud-hosted deployment (subscriptions ranging from $20 to $200 monthly). It supports web search, code generation, image analysis, and remote control via Telegram and Discord. The company also advanced DisTrO, its distributed training framework, reducing communication overhead across geographically dispersed GPUs, and continued developing Psyche Network, a Solana-based coordination layer for decentralized compute, data, and governance.
Looking ahead, Nous Research plans to allocate new capital toward commercializing Hermes (enterprise controls, audit logs, integrations), scaling DisTrO infrastructure for fault-tolerant training on heterogeneous GPU networks, and refining Psyche's tokenomics and governance to incentivize compute providers and contributors. The company emphasizes a dual strategy: maintaining open-source credibility while expanding paid hosted services, positioning itself as a bridge between Web3 coordination and practical AI deployment.
Nous Research was co-founded in 2023 by Jeffrey Quesnelle, Karan Malhotra (Mephisto), Teknium, and Shivani Mitra.
Jeffrey Quesnelle: Co-founder & CTO. Lead author of the YaRN paper, overseeing technical strategy and research.
Karan Malhotra (Mephisto): Co-founder & Head of Behavior. Oversees communications and strategy for Hermes, WorldSim, etc.
Teknium: Co-founder & Head of Post Training. Core developer of the Hermes model family.
Dillon Rolnick: CEO. Responsible for day-to-day operations and overall management.