Hermes 4 405B reads text and writes text back, and it keeps about 128 pages in mind at once. Built on Llama 3.1 405B, it can think out loud before answering or reply directly, depending on a setting. Its extra training, heavy on worked reasoning, aims at math, code, science and logic, and it fills in exact forms, calls the functions you connect and refuses less. Hermes 3 405B is its predecessor on the same base.
Hermes 4 70B
by NousresearchReleased 26 August 2025
Hermes 4 70B is a hybrid reasoning model from Nous Research, built on Meta-Llama-3.1-70B. It introduces the same hybrid mode as the larger 405B release, allowing the model to either...
Reads about 130 pages · writes up to 110 pages
Parameters you can send
Standard OpenAI fields, grouped by what they control. Anything not listed is accepted and ignored.
Shaping the reply
How long, how varied, where to stop.
max_tokenstemperaturetop_ptop_kfrequency_penaltypresence_penaltyrepetition_penaltyReasoning
Let it think before answering — and see the thinking.
reasoninginclude_reasoningStructured output
Get JSON you can parse instead of prose.
response_formatQuestions about Hermes 4 70B
Is Hermes 4 70B still on sale?
No. Nousresearch has retired it. Other Nousresearch models are on sale.
What is Hermes 4 70B best for?
The lab has not said.
Does Hermes 4 70B work in Arabic, French or Darja?
Nousresearch names no languages for Hermes 4 70B. Arabic, French and Darja: not stated.
Can it read PDFs, images or a whole book?
It reads text. Up to 131K tokens in one request, about 130 pages. It writes up to 88 000 words at once.
Is Hermes 4 70B open source?
Yes, its weights are public. Open weights on Hugging Face
Who makes Hermes 4 70B, and when was it released?
Nousresearch. Released 26 August 2025. It knows the world up to 31 August 2024.