Open AI models and agent software
Nous Research
Nous Research trains AI models and builds open-source agent software for developers, researchers and people who want control over their AI tools. Formed in 2023 to challenge centralized access to AI, the company initially concentrated on post-training: adapting existing models to follow users’ instructions and preferences. Its work has since expanded from fine-tuning Meta’s Llama models to building Hermes Agent, which lets users choose both the models they use and where their agent runs.
The company’s research connects model behavior with the practical limits of using models. Its early YaRN work extended open models’ context windows at lower additional cost, allowing them to process more of a conversation, document or codebase at once. The Hermes 3 models, introduced in August 2024, focused on instruction following and tool use alongside reasoning and creative tasks. This work reflects Nous’s emphasis on adapting models for interaction, rather than treating a pretrained model as a finished assistant.
Hermes Agent carries that emphasis into software that retains knowledge across sessions. Its learning loop creates reusable skills from complex tasks and refines them during subsequent use; conversation search helps it recover earlier work. Users can run the agent locally or on remote infrastructure and interact through a terminal or messaging services. Model choice remains separate from the agent’s accumulated skills and memory, so users can change providers without replacing the whole setup. The code is available under the MIT license, giving developers a foundation they can inspect and modify.
1 talk
Newest firstLoading more talks and speakers…
1 speaker at AIE
Affiliations reflect their AIE event listings and recorded appearances, not necessarily current employment.
