Models_
Hermes 4 is a frontier, hybrid-mode reasoning model. It extends Hermes 3 with stronger math and science reasoning, better instruction following and schema-adherent outputs, and more nuanced roleplay and writing.
Hybrid thinking models require the following system prompt to activate reasoning mode:
You are a deep thinking AI, you may use extremely long chains of thought to deeply consider the problem and deliberate with yourself via systematic reasoning processes to help come to a correct solution prior to answering. You should enclose your thoughts and internal monologue inside <think> </think> tags, and then provide your solution or response to the problem.
Hermes-4-70B
Model CardAvailable to all API users
Was: $0.70/1M tokens
$0.05/1M prompt tokens
$0.20/1M completion tokens
Try for free on Nous Chat.
128k tokens ctx
Thinking:
hybrid
This incarnation of Hermes 4 balances scale and size. It handles complex reasoning tasks, while staying fast and cost effective. A versatile choice for many use cases.
This model is based on Llama-3.1-70B.
Hermes-4-405B
Model CardAvailable to all API users
Was: $1.50/1M tokens
$0.09/1M prompt tokens
$0.37/1M completion tokens
Try for free on Nous Chat.
128k tokens ctx
Thinking:
hybrid
This is the largest model in the Hermes 4 family, and it is the fullest expression of our design, focused on advanced reasoning and creative depth rather than optimizing inference speed or cost.
This model is based on Llama-3.1-405B.
3rd Party Models
Subscribers using Nous Chat and Hermes Agent can also access the following models (powered by OpenRouter):
| Model | Token Price |
|---|---|
| claude-haiku-4-5-20251001 | in $1.10 / out $5.50 per 1M |
| claude-opus-4-6 | in $5.50 / out $27.50 per 1M |
| claude-sonnet-4-6 | in $3.30 / out $16.50 per 1M |
| deepseek-r1 | in $0.77 / out $2.75 per 1M |
| deepseek-v3.2 | in $0.28 / out $0.45 per 1M |
| deepseek/deepseek-v3.2 | in $0.28 / out $0.45 per 1M |
| gemini-3-flash | in $0.55 / out $3.30 per 1M |
| gemini-3.0-pro-preview |
Unsupported Models
Hermes-4.3-36B
Model CardCurrently unavailable via the API
You can use this model locally, but if you're using our hosted API we recommend migrating to Hermes-4-70B.
N/A
128k tokens ctx
Thinking:
hybrid
Hermes 4.3 is optimised for local deployment. You can dowload it from Hugging Face for local use.