Nous Research

Nous Research is an artificial intelligence company specializing in open-source language models and decentralized training infrastructure. Their flagship product, Hermes, is a series of general-purpose, human-aligned models fine-tuned on extensive datasets to enhance performance across various tasks. The company also developed DiStrO (Distributed Training Over-the-Internet), a technology enabling efficient data compression between globally distributed GPUs, reducing training costs and improving scalability. Additionally, they introduced the Psyche Network, built on the Solana blockchain, facilitating collaborative AI training while resisting censorship.

Founded in 2023 by Karan Malhotra, Shivani Mitra, Jeffrey Quesnelle, and Teknium, Nous Research has achieved significant milestones. In January 2024, they secured $5.2 million in seed funding, followed by a $50 million Series A round in 2025, primarily led by Paradigm Operations, valuing the company at $1 billion. These investments underscore the growing interest in decentralized AI development and Nous Research's role in this sector.

The Hermes series includes models like Hermes 3, Hermes 4, and Hermes 4.3, each offering advancements in context length, reasoning capabilities, and alignment with user values. For instance, Hermes 4.3, trained with an extended context length of up to 512K, matches or exceeds the performance of Hermes 4 70B at half the parameter cost. These models are available on platforms like Hugging Face, facilitating widespread adoption and further development.

Nous Research's customer base spans AI researchers, developers, and organizations seeking open-source AI solutions. Their partnerships with entities like OpenAI co-founder Diedrick Kingma have contributed to the creation of models used by companies such as Meta and DeepSeek. The company's market position is strengthened by its focus on decentralized AI, distinguishing it from competitors like OpenAI and Anthropic, and its commitment to open-source development fosters a collaborative approach to AI advancement.