Nous Research Releases Hermes 4 Open-Source AI Model Family
The new models, released August 30, are built on Llama 3.1 and feature a "hybrid reasoning" mode with minimal content restrictions.
Nous Research on August 30, 2025, unveiled its Hermes 4 family of open-source large language models, which it claims rivals the performance of leading proprietary systems like GPT-4o and Claude Sonnet 4. The models are designed to provide users with greater control and fewer content restrictions, a direct challenge to the "safety-filtered" approach of major AI labs.
A Challenge to Proprietary Systems
The Hermes 4 family consists of models with 14 billion, 70 billion, and 405 billion parameters, all built on Meta's Llama 3.1 architecture. The company emphasized that the models are designed with "neutral alignment" and minimal content …
Archive Access
This article is older than 24 hours. Create a free account to access our 7-day archive.