AI Desk · Model Identity · Compare

Meta: Llama 4 Maverick · Meta: Muse Glimmer 30B

Identity similarity: highNear-exact agreement · 48/50Agreement this close usually means one tokenizer behind two serving templates, or a small vocabulary revision inside one family. The differing rows below say which strings part.

Meta: Llama 4 Maverick recordMeta: Muse Glimmer 30B recordOpen in the interactive bench

Tokenizer fingerprint

48 of 50 mutually clean strings agree
Test stringMeta: Llama 4 MaverickMeta: Muse Glimmer 30BVerdict
en-prose1414match
en-long1717match
spaces-2033match
spaces-6033match
tabs-2033match
newlines-2044match
mixed-ws66match
digits-933match
digits-1244match
digits-301010match
digits-sep77match
float-long99match
zh-common1111match
zh-long1212match
zh-rare1818match
ja-kana99match
ja-kanji1212match
ko1010match
ru1010match
ar1111match
he1010match
hi1212match
th88match
el1414match
emoji-basic1010match
emoji-skin2020match
emoji-zwj-family1111match
emoji-zwj-x33333match
emoji-flags1616match
emoji-prof1919match
math2424match
boxdraw2121match
combining1010match
cjk-ext-b1616match
surrogates3737match
zalgo3131match
rtl-mix66match
py-code2525match
py-indent1515match
json2020match
html1616match
regex4242match
camel66match
snake66match
rare-word-x53031differs
repeat-tok4041differs
base643535match
hex1111match
url1717match
uuid2727match

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
FieldMeta: Llama 4 MaverickMeta: Muse Glimmer 30BVerdict
Context length1048576131072× differs
Max output tokens16384131072× differs
Supported parametersfrequency_penalty, logit_bias, logprobs, max_tokens, min_p, presence_penalty, repetition_penalty, response_format, seed, stop, structured_outputs, temperature, tool_choice, tools, top_k, top_logprobs, top_pfrequency_penalty, include_reasoning, logit_bias, logprobs, max_tokens, min_p, presence_penalty, reasoning, reasoning_effort, repetition_penalty, response_format, seed, stop, structured_outputs, temperature, tool_choice, tools, top_k, top_logprobs, top_p× differs
Defaults{}{"temperature":1,"top_p":0.95,"top_k":64}× differs
Reasoning contract{"mandatory":true,"supported_efforts":["xhigh","high","medium","low"],"default_effort":"medium"}× differs
Modalitiestext+image->texttext+image->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.