Nous Research Glossary
Nous Research is a decentralised open-source AI research collective founded in 2023 by a small group of independent researchers including Karan Malhotra, Bowen Peng, and Teknium (a pseudonymous figure who has been one of the most visible open-source-AI contributors of the era). The collective is best known for the Hermes family of fine-tuned models — instruction-following adaptations of open-weights base models (typically Llama) — and for a sustained body of post-training research that has been influential well beyond the collective’s own releases.
For this Dictionary, Nous Research matters as one of the most consequential independent open-weights labs in the 2025–2026 ecosystem. Where Mistral and Meta release base models, Nous fine-tunes and post-trains those base models into something more directly useful for operators running local inference. Hermes is, in practice, the default candidate the Dictionary’s Sovereign Compute calculator recommends for operators who want a known-quantity instruction-tuned local model without doing the fine-tuning themselves.
The collective’s funding model (a mix of grants, independent compute, and occasional commercial partnerships) is itself a Dictionary-relevant data point: a working alternative to the lab-as-startup paradigm.
See also
- Hermes — the model family
- Llama — the typical base
- Sovereign Compute
- Hugging Face — Nous’s primary distribution channel