AI Desk · Model Identity · Compare
Nous: Hermes 4 70B · Sao10K: Llama 3 8B Lunaris
Identity similarity: lowDivergent fingerprints · 12/50These models tokenise differently. Divergence on the high-information strings (CJK, emoji, digits) separates unrelated vocabularies within a handful of samples.
Nous: Hermes 4 70B recordSao10K: Llama 3 8B Lunaris recordOpen in the interactive bench
Tokenizer fingerprint
12 of 50 mutually clean strings agree| Test string | Nous: Hermes 4 70B | Sao10K: Llama 3 8B Lunaris | Verdict |
|---|---|---|---|
| en-prose | 14 | 13 | differs |
| en-long | 16 | 15 | differs |
| spaces-20 | 3 | 2 | differs |
| spaces-60 | 3 | 3 | match |
| tabs-20 | 3 | 2 | differs |
| newlines-20 | 4 | 4 | match |
| mixed-ws | 5 | 4 | differs |
| digits-9 | 3 | 2 | differs |
| digits-12 | 4 | 3 | differs |
| digits-30 | 10 | 10 | match |
| digits-sep | 7 | 6 | differs |
| float-long | 9 | 8 | differs |
| zh-common | 12 | 11 | differs |
| zh-long | 18 | 18 | match |
| zh-rare | 18 | 18 | match |
| ja-kana | 11 | 10 | differs |
| ja-kanji | 12 | 11 | differs |
| ko | 14 | 13 | differs |
| ru | 14 | 13 | differs |
| ar | 12 | 11 | differs |
| he | 24 | 23 | differs |
| hi | 15 | 14 | differs |
| th | 14 | 14 | match |
| el | 13 | 12 | differs |
| emoji-basic | 10 | 9 | differs |
| emoji-skin | 30 | 29 | differs |
| emoji-zwj-family | 15 | 15 | match |
| emoji-zwj-x3 | 45 | 44 | differs |
| emoji-flags | 24 | 24 | match |
| emoji-prof | 26 | 25 | differs |
| math | 31 | 30 | differs |
| boxdraw | 23 | 22 | differs |
| combining | 11 | 10 | differs |
| cjk-ext-b | 13 | 13 | match |
| surrogates | 40 | 39 | differs |
| zalgo | 35 | 34 | differs |
| rtl-mix | 8 | 7 | differs |
| py-code | 25 | 24 | differs |
| py-indent | 15 | 14 | differs |
| json | 20 | 19 | differs |
| html | 16 | 15 | differs |
| regex | 44 | 44 | match |
| camel | 5 | 4 | differs |
| snake | 6 | 6 | match |
| rare-word-x5 | 31 | 29 | differs |
| repeat-tok | 22 | 20 | differs |
| base64 | 35 | 34 | differs |
| hex | 11 | 11 | match |
| url | 17 | 16 | differs |
| uuid | 27 | 26 | differs |
Values are the marginal prompt-token cost of each string with the model's fixed overhead subtracted. An x marks a row excluded as corrupt for that model.
API surface
The declared serving contracts differ| Field | Nous: Hermes 4 70B | Sao10K: Llama 3 8B Lunaris | Verdict |
|---|---|---|---|
| Context length | 131072 | 8192 | × differs |
| Max output tokens | – | 16384 | × differs |
| Supported parameters | frequency_penalty, include_reasoning, max_tokens, presence_penalty, reasoning, repetition_penalty, response_format, temperature, top_k, top_p | frequency_penalty, logit_bias, logprobs, max_tokens, min_p, presence_penalty, repetition_penalty, response_format, seed, stop, structured_outputs, temperature, top_k, top_logprobs, top_p | × differs |
| Defaults | {} | {} | ✓ match |
| Reasoning contract | {"mandatory":false} | – | × differs |
| Modalities | text->text | text->text | ✓ match |
Method: tokenizer fingerprinting, v1, 50 strings. Raw data on the data page. This page is static and dated; the interactive bench compares any two of the 422 catalogue models.