AI Desk · Model Identity · Compare

Perceptron: Perceptron Mk1 · Tencent: Hunyuan A13B Instruct

Identity similarity: lowDivergent fingerprints · 30/50These models tokenise differently. Divergence on the high-information strings (CJK, emoji, digits) separates unrelated vocabularies within a handful of samples.

Perceptron: Perceptron Mk1 recordTencent: Hunyuan A13B Instruct recordOpen in the interactive bench

Tokenizer fingerprint

30 of 50 mutually clean strings agree
Test stringPerceptron: Perceptron Mk1Tencent: Hunyuan A13B InstructVerdict
en-prose1414match
en-long1616match
spaces-2033match
spaces-6033match
tabs-2033match
newlines-2044match
mixed-ws55match
digits-999match
digits-121212match
digits-303030match
digits-sep1212match
float-long2222match
zh-common99match
zh-long1010match
zh-rare1818match
ja-kana713differs
ja-kanji1013differs
ko1319differs
ru1216differs
ar1025differs
he1424differs
hi1532differs
th825differs
el1328differs
emoji-basic910differs
emoji-skin3019differs
emoji-zwj-family1811differs
emoji-zwj-x35433differs
emoji-flags2424match
emoji-prof2922differs
math2929match
boxdraw2826differs
combining56differs
cjk-ext-b1313match
surrogates4040match
zalgo3635differs
rtl-mix611differs
py-code2625differs
py-indent1815differs
json2525match
html1616match
regex4444match
camel55match
snake66match
rare-word-x53131match
repeat-tok2222match
base643636match
hex1717match
url1919match
uuid3636match

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
FieldPerceptron: Perceptron Mk1Tencent: Hunyuan A13B InstructVerdict
Context length32768131072× differs
Max output tokens8192131072× differs
Supported parametersfrequency_penalty, include_reasoning, max_tokens, presence_penalty, reasoning, structured_outputs, temperature, top_k, top_pfrequency_penalty, include_reasoning, max_tokens, reasoning, response_format, structured_outputs, temperature, top_k, top_p× differs
Defaults{"temperature":null,"top_p":null,"top_k":null,"frequency_penalty":null,"presence_penalty":null,"repetition_penalty":null}{}× differs
Reasoning contract{"mandatory":false}{"mandatory":false}✓ match
Modalitiestext+image+video->texttext->text× differs

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.