Hermes
Family overview — summarizes releases; licenses and availability belong to each release.
Maintained by Nous Research12
Hermes is Nous Research's series of post-trained language models, each built on a base model released by another developer. Releases listed by Nous include Nous-Hermes Llama 2 13B (July 2023), Hermes 2 models on Mistral, Mixtral, Yi, and Llama 3 bases (2023 to 2024), Hermes 3 on Llama 3.1 (August 2024), Hermes 4 at 405B, 70B, and 14B (August 2025), and Hermes 4.3 36B (December 2025). Hermes 4 and 4.3 are hybrid reasoning models that can answer directly or reason in think tags first.1523
- Release notes: Nous Research releases (external site: nousresearch.com)
- Paper: Hermes 4 technical report (arXiv 2508.18255) (external site: arxiv.org)
- Release notes: Hermes 4.3 announcement (external site: nousresearch.com)
Releases assessed
| Release | Weights | Tier (USASI rubric v0.1) | License | Released |
|---|---|---|---|---|
| Hermes 4 70B | Weights: Public | Model-disclosure tier (USASI rubric v0.1): Open-weight | Meta Llama 3 Community License (card license field "llama3") | Aug 26, 2025 |
| Hermes 4.3 36B | Weights: Public | Model-disclosure tier (USASI rubric v0.1): Open-weight | Apache-2.0 | Dec 3, 2025 |
What it is useful for
Organization context
Provenance and derivatives
Every Hermes release is a fine-tune of another developer's base model. Hermes 4 405B and 70B start from Meta's Llama 3.1 405B and 70B, Hermes 4 14B from Qwen3 14B, and Hermes 4.3 36B from ByteDance Seed's Seed-OSS-36B-Base. Earlier generations used Llama 2, Mistral 7B, Mixtral 8x7B, Yi 34B, Llama 3, and Llama 3.1 bases. Nous Research's own work is the post-training data, the fine-tuning, and (for Hermes 4.3) decentralized training on Psyche.231576
- Derived from: Llama 3.1 (405B and 70B base) — Meta base models for Hermes 3 and Hermes 4 405B/70B.
- Derived from: Qwen3-14B (external site: huggingface.co) — Qwen team base model for Hermes 4 14B; not a catalog member.
- Derived from: Seed-OSS-36B-Base (external site: huggingface.co) — ByteDance Seed Team base model for Hermes 4.3 36B; not a catalog member.
U.S. eligibility
Eligible · basis: U.S. headquarters
The Hermes models are post-trained and published by Nous Research, Inc., a Delaware corporation with U.S. addresses (see the nous-research record). Eligibility rests on Nous Research as maintainer of the fine-tunes; it does not extend to the base models, which were developed by other organizations. Each release is assessed on its own record.892
Sources
This listing is not an endorsement, a safety assessment, or a federal approval.