YFarmXIntelligence for an accelerating worldBriefings
YFarmX
YFarmX

AI Desk · Model Identity

Qwen: Qwen3 VL 235B A22B Thinking

qwen/qwen3-vl-235b-a22b-thinking · Alibaba (Qwen)Declared and measured agree
Tokenizer lineage (measured)
Qwen
Evidence confidence
High
Why high: an exact fingerprint shared with 36 models that declare the same family, without a second signal of a different kind yet.
Declared tokenizer tag
Qwen3
Entered catalogue
23 September 2025
Last tested
3 September 2026
Measurement
Measured
Evidence
  • ✓ exact tokenizer signature shared with 36 models
Compare with Qwen: Qwen3 Max Thinking

Download JSONReport a correction

Identity statement

Qwen: Qwen3 VL 235B A22B Thinking is Alibaba (Qwen)'s model. Its measured tokenizer sits with the Qwen group, which is evidence that it builds on the Qwen tokenizer. Tokenizer reuse is normal engineering, open vocabularies travel between labs, and this says nothing against Alibaba (Qwen)'s authorship of the model itself.

Evidence stack

Signals of different kinds, weighed together
SignalResultReference
Tokenizer signatureExact match with 36 other modelstk_5410b38e
API surfaceNo other entry shares this exact contractapi_76b57a1b
Context and output131,072 context · 32,768 max outputdeclared
Reasoning contractMandatorydeclared
Serving providers observedAlibaba, Novitameasured 3 September 2026

Shares this fingerprint

Exact signature first; template-boundary shifts of the same signature beneath
Magnum v4 72Banthracite-org/magnum-v4-72bexact · 50/50ByteDance: UI-TARS 7B bytedance/ui-tars-1.5-7bexact · 50/50Dots Studio: Dots3-Note Preview (free)dots-studio/dots-3-note-preview:freeexact · 50/50Morph: Morph V3 Fastmorph/morph-v3-fastexact · 50/50Qwen2.5 72B Instructqwen/qwen-2.5-72b-instructexact · 50/50Qwen: Qwen2.5 7B Instructqwen/qwen-2.5-7b-instructexact · 50/50Qwen2.5 Coder 32B Instructqwen/qwen-2.5-coder-32b-instructexact · 50/50Qwen: Qwen-Plusqwen/qwen-plusexact · 50/50Qwen: Qwen Plus 0728qwen/qwen-plus-2025-07-28exact · 50/50Qwen: Qwen2.5 VL 72B Instructqwen/qwen2.5-vl-72b-instructexact · 50/50Qwen: Qwen3 14Bqwen/qwen3-14bexact · 50/50Qwen: Qwen3 235B A22Bqwen/qwen3-235b-a22bexact · 50/50Qwen: Qwen3 235B A22B Instruct 2507qwen/qwen3-235b-a22b-2507exact · 50/50Qwen: Qwen3 235B A22B Thinking 2507qwen/qwen3-235b-a22b-thinking-2507exact · 50/50Qwen: Qwen3 30B A3Bqwen/qwen3-30b-a3bexact · 50/50Qwen: Qwen3 30B A3B Instruct 2507qwen/qwen3-30b-a3b-instruct-2507exact · 50/50Qwen: Qwen3 30B A3B Thinking 2507qwen/qwen3-30b-a3b-thinking-2507exact · 50/50Qwen: Qwen3 32Bqwen/qwen3-32bexact · 50/50Qwen: Qwen3 8Bqwen/qwen3-8bexact · 50/50Qwen: Qwen3 Coder 480B A35Bqwen/qwen3-coderexact · 50/50Qwen: Qwen3 Coder 30B A3B Instructqwen/qwen3-coder-30b-a3b-instructexact · 50/50Qwen: Qwen3 Coder Flashqwen/qwen3-coder-flashexact · 50/50Qwen: Qwen3 Coder Nextqwen/qwen3-coder-nextexact · 50/50Qwen: Qwen3 Coder Plusqwen/qwen3-coder-plusexact · 50/50Qwen: Qwen3 Maxqwen/qwen3-maxexact · 50/50Qwen: Qwen3 Max Thinkingqwen/qwen3-max-thinkingexact · 50/50Qwen: Qwen3 Next 80B A3B Instructqwen/qwen3-next-80b-a3b-instructexact · 50/50Qwen: Qwen3 Next 80B A3B Thinkingqwen/qwen3-next-80b-a3b-thinkingexact · 50/50Qwen: Qwen3 VL 235B A22B Instructqwen/qwen3-vl-235b-a22b-instructexact · 50/50Qwen: Qwen3 VL 30B A3B Instructqwen/qwen3-vl-30b-a3b-instructexact · 50/50Qwen: Qwen3 VL 30B A3B Thinkingqwen/qwen3-vl-30b-a3b-thinkingexact · 50/50Qwen: Qwen3 VL 32B Instructqwen/qwen3-vl-32b-instructexact · 50/50Qwen: Qwen3 VL 8B Instructqwen/qwen3-vl-8b-instructexact · 50/50Qwen: Qwen3 VL 8B Thinkingqwen/qwen3-vl-8b-thinkingexact · 50/50Relace: Relace Searchrelace/relace-searchexact · 50/50Xiaomi: MiMo-V2.5xiaomi/mimo-v2.5exact · 50/50Xiaomi: MiMo-V2.5-Proxiaomi/mimo-v2.5-prosame tokenizer · template shiftWriter: Palmyra X5writer/palmyra-x5same tokenizer · template shift

Closest measured models

Agreement across the mutually clean test strings
Xiaomi: MiMo-V2.5xiaomi/mimo-v2.550/50 · 100%Dots Studio: Dots3-Note Preview (free)dots-studio/dots-3-note-preview:free50/50 · 100%Qwen: Qwen3 30B A3B Instruct 2507qwen/qwen3-30b-a3b-instruct-250750/50 · 100%Qwen: Qwen3 Coder 30B A3B Instructqwen/qwen3-coder-30b-a3b-instruct50/50 · 100%Qwen: Qwen3 32Bqwen/qwen3-32b50/50 · 100%ByteDance: UI-TARS 7B bytedance/ui-tars-1.5-7b50/50 · 100%

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-99
digits-1212
digits-3030
digits-sep12
float-long22

CJK

zh-common10
zh-long12
zh-rare17
ja-kana8
ja-kanji13
ko17

Other scripts

ru16
ar8
he9
hi29
th12
el28

Emoji

emoji-basic5
emoji-skin10
emoji-zwj-family10
emoji-zwj-x330
emoji-flags8
emoji-prof14

Rare Unicode

math26
boxdraw17
combining5
cjk-ext-b12
surrogates22
zalgo36
rtl-mix7

Code

py-code25
py-indent15
json25
html16
regex44
camel5
snake6

Repetition and encodings

rare-word-x531
repeat-tok22
base6436
hex17
url19
uuid36

An x marks a row excluded as corrupt (caching or provider interference detected during measurement). Overhead subtracted: 10 prompt tokens.

Declared record

What the catalogue claims about this model

Qwen3-VL-235B-A22B Thinking is a multimodal model that unifies strong text generation with visual understanding across images and video. The Thinking model is optimized for multimodal reasoning in STEM and math....

Modalities
text+image->text
Declared tokenizer
Qwen3
Prompt price
$0.40 / M tokens
Completion price
$4.00 / M tokens

Supported parameters

frequency_penaltyinclude_reasoninglogprobsmax_tokenspresence_penaltyreasoningrepetition_penaltyresponse_formatseedstopstructured_outputstemperaturetool_choicetoolstop_ktop_logprobstop_p

defaults: {"temperature":0.8,"top_p":0.95,"top_k":20,"frequency_penalty":null,"presence_penalty":null,"repetition_penalty":1}

Catalogue entryWeights on Hugging Face

History

Every observation, kept as taken
  • 23 September 2025Enters the OpenRouter catalogue declared as Qwen3.
  • 3 September 2026Fingerprinted in the YFarmX catalogue sweep · 50 of 50 strings measured clean.
  • 3 September 2026YFarmX assessment: consistent with the Qwen family, high confidence.

Cite this page as the evidence record for Qwen: Qwen3 VL 235B A22B Thinking: the URL is stable, measurements are dated, and revisions append to the history above. Method: tokenizer fingerprinting, v1, 50 strings.