Z.ai: GLM 5.3

z-ai/glm-5.3 · Z.aiTokenizer lineage measured

Tokenizer group (measured): Z.ai
Evidence confidence: Very high
Declared tokenizer tag: Other
Entered catalogue: 18 August 2026
Last tested: 21 August 2026
Measurement: Measured

Evidence

exact tokenizer signature shared with 2 models
API-surface contract identical to a group member

Compare with Z.ai: GLM 4.7 Flash

Download JSON Report a correction

Identity statement

Z.ai: GLM 5.3 is Z.ai's model, and its measured tokenizer sits in Z.ai's own signature group alongside its sibling models. Measured models from Undisclosed (stealth) carry the same vocabulary, which reads as reuse of Z.ai's tokenizer line and questions nobody's authorship; they are listed under "Shares this fingerprint" below.

Evidence stack

Signals of different kinds, weighed together

Signal	Result	Reference
Tokenizer signature	Exact match with 2 other models	tk_133b086c
API surface	Identical contract to 2 other entries	api_7e146681
Context and output	1,048,576 context · 131,072 max output	declared
Reasoning contract	Mandatory · efforts max, high, low · default max	declared
Serving providers observed	Z.AI	measured 21 August 2026

Shares this fingerprint

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

Ox Alphastealth/ox-alphaexact · 50/50 Z.ai: GLM 4.7 Flashz-ai/glm-4.7-flashexact · 50/50 Z.ai: GLM 5.2z-ai/glm-5.2same tokenizer · template shift

Closest measured models

Agreement across the mutually clean test strings

Ox Alphastealth/ox-alpha50/50 · 100%Z.ai: GLM 4.7 Flashz-ai/glm-4.7-flash50/50 · 100%Z.ai: GLM 5.2z-ai/glm-5.249/50 · 98%Z.ai: GLM 4.6Vz-ai/glm-4.6v40/50 · 80%Z.ai: GLM 4.5 Airz-ai/glm-4.5-air40/50 · 80%Meta: Llama 3.2 1B Instructmeta-llama/llama-3.2-1b-instruct24/50 · 48%

Click a row to open the full comparison.

The measured fingerprint

Marginal prompt-token cost of each test string, grouped by script

All 50 rows

English and whitespace

en-prose14

en-long16

spaces-203

spaces-603

tabs-203

newlines-204

mixed-ws5

Digits

digits-95

digits-127

digits-3010

digits-sep9

float-long14

CJK

zh-common9

zh-long10

zh-rare18

ja-kana10

ja-kanji14

ko17

Other scripts

ru13

ar13

he24

hi27

th24

el14

Emoji

emoji-basic5

emoji-skin10

emoji-zwj-family7

emoji-zwj-x321

emoji-flags24

emoji-prof15

Rare Unicode

math27

boxdraw28

combining10

cjk-ext-b13

surrogates40

zalgo22

rtl-mix8

Code

py-code25

py-indent15

json21

html16

regex44

camel5

snake6

Repetition and encodings

rare-word-x531

repeat-tok22

base6435

hex14

url17

uuid28

An x marks a row excluded as corrupt (caching or provider interference detected during measurement). Overhead subtracted: 12 prompt tokens.

Declared record

What the catalogue claims about this model

GLM-5.3 is a large-scale reasoning model from Z.ai, built for complex software engineering and long-horizon agent tasks. It supports text input and output with a 1M-token context window, and improves...

Modalities: text->text
Declared tokenizer: Other
Prompt price: $1.40 / M tokens
Completion price: $4.40 / M tokens

Supported parameters

include_reasoningmax_tokensreasoningreasoning_effortresponse_formattemperaturetool_choicetoolstop_ktop_p

defaults: {"temperature":1,"top_p":0.95}

Catalogue entry

History

Every observation, kept as taken

18 August 2026Enters the OpenRouter catalogue with no declared family.
20 August 2026Fingerprinted during the Ox Alpha investigation · 50 of 50 strings measured clean.
21 August 2026Fingerprinted in the YFarmX catalogue sweep · 50 of 50 strings measured clean.
21 August 2026YFarmX assessment: consistent with the Z.ai family, very high confidence.

Cite this page as the evidence record for Z.ai: GLM 5.3: the URL is stable, measurements are dated, and revisions append to the history above. Method: tokenizer fingerprinting, v1, 50 strings. The raw data behind every figure is on the data page.