Nous: Hermes 4 405B

nousresearch/hermes-4-405b · Nous ResearchTokenizer lineage measured

Tokenizer lineage (measured): Llama3
Entered catalogue: 26 August 2025
Last tested: 3 September 2026
Evidence confidence: Very high

Why very high: an exact fingerprint shared with 4 models and an identical API contract inside the same group: two signals of different kinds agree.

Evidence

✓ exact tokenizer signature shared with 4 models
✓ API-surface contract identical to a group member

Compare with Nous: Hermes 4 70B

Report a correction

Identity statement

Nous: Hermes 4 405B is Nous Research's model. Its measured tokenizer sits with the Llama3 group, which is evidence that it builds on the Llama3 tokenizer. Tokenizer reuse is normal engineering, open vocabularies travel between labs, and this says nothing against Nous Research's authorship of the model itself.

What we checked

Four different kinds of evidence, and what each one showed

We checked	What we found	Where it came from
How it counts tokens	Counts every one of the 50 test strings exactly like 4 other models	tk_47615f2d
How its API is set up	Set up identically to 1 other listing, field for field	api_9f8a0c50
How much it can read and write	Reads up to 131,072 tokens at onceOpenRouter's listing shows 117,964 output tokens, which is 90% of the context window, the figure it carries where the provider declares no maximum.	its own listing
Whether it thinks before answering	Reasons when asked to	its own listing
Who served our requests	Nebius	we measured it, 3 September 2026

Shares this fingerprint

Exact signature first; template-boundary shifts of the same signature beneath

Closest measured models

Agreement across the mutually clean test strings

Meta: Llama 3.2 1B Instructmeta-llama/llama-3.2-1b-instruct50/50 · 100%Nous: Hermes 4 70Bnousresearch/hermes-4-70b50/50 · 100%Nous: Hermes 3 70B Instructnousresearch/hermes-3-llama-3.1-70b50/50 · 100%Nous: Hermes 3 405B Instructnousresearch/hermes-3-llama-3.1-405b50/50 · 100%Meta: Llama 3.2 3B Instructmeta-llama/llama-3.2-3b-instruct48/50 · 96%Meta: Llama 3.1 8B Instructmeta-llama/llama-3.1-8b-instruct48/50 · 96%

Click a row to open the full comparison.

The measured fingerprint

What each test string cost this model, in tokens

We sent Nous: Hermes 4 405B fifty short pieces of text and recorded what each one cost it in tokens. The bars below are those costs. Two models built on the same tokenizer produce the same bars; a model built on a different one produces a different set, which is what makes this a fingerprint.

All 50 rows

English and whitespace

en-prose14

en-long16

spaces-203

spaces-603

tabs-203

newlines-204

mixed-ws5

Digits

digits-93

digits-124

digits-3010

digits-sep7

float-long9

CJK

zh-common12

zh-long18

zh-rare18

ja-kana11

ja-kanji12

ko14

Other scripts

ru14

ar12

he24

hi15

th14

el13

Emoji

emoji-basic10

emoji-skin30

emoji-zwj-family15

emoji-zwj-x345

emoji-flags24

emoji-prof26

Rare Unicode

math31

boxdraw23

combining11

cjk-ext-b13

surrogates40

zalgo35

rtl-mix8

Code

py-code25

py-indent15

json20

html16

regex44

camel5

snake6

Repetition and encodings

rare-word-x531

repeat-tok22

base6435

hex11

url17

uuid27

Overhead subtracted: 24 prompt tokens.

Declared record

What the catalogue claims about this model

Hermes 4 is a large-scale reasoning model built on Meta-Llama-3.1-405B and released by Nous Research. It introduces a hybrid reasoning mode, where the model can choose to deliberate internally with...

Modalities: text->text
Declared tokenizer: Other
Prompt price: $1.00 / M tokens
Completion price: $3.00 / M tokens

Supported parameters

frequency_penaltyinclude_reasoningmax_tokenspresence_penaltyreasoningrepetition_penaltyresponse_formattemperaturetop_ktop_p

defaults: {}

Catalogue entry Weights on Hugging Face

History

Every observation, kept as taken

26 August 2025Enters the OpenRouter catalogue with no declared family.
3 September 2026Fingerprinted in the YFarmX catalogue sweep · 50 of 50 strings measured clean.
3 September 2026YFarmX assessment: consistent with the Llama3 family, very high confidence.