Data Explorer

Per-model self-identification rates across languages, grouped by vendor. 85 cross-vendor confusions detected.

2,600 answers·26 models·10 languages

Run

Run summary

26 models · 10 languages

2,600

Answers

88.8%

Self-ID

3.3%

Cross-vendor

7.1%

Unknown

0.8%

Refused

Imitation balance

Per manufacturer: how often other models claim to be it (right) vs how often its own models claim to be someone else (left). Sorted by net — identity creditors at the top, debtors at the bottom. Tested manufacturers only.

DeepSeek

+22

OpenAI

+17

Qwen

Google

Xiaomi

Moonshot

xAI

inclusionAI

z-ai

ERNIE

MiniMax

-1

Anthropic

-2

StepFun

-3

Kwai

-9

Doubao

-18

Tencent

-45

imitates others ◀│▶ imitated by others

Strongest confusion pairs

The most likely directed mistakes — when a manufacturer's models claim to be a specific other manufacturer. Bar length = how often (P across all answers for the source).

Tencent→

DeepSeek

15.0%15/100

Tencent→Microsoft

12.0%12/100

Tencent→

Anthropic

7.0%7/100

Kwai→

Qwen

6.0%6/100

Doubao→

OpenAI

3.3%13/400

Anthropic→

DeepSeek

2.0%6/300

Tencent→Naver

2.0%2/100

StepFun→

Google

2.0%2/100

Doubao→Yandex

1.3%5/400

MiniMax→

Xiaomi

1.0%1/100

Tencent→Ha-an

1.0%1/100

Tencent→SK Telecom

1.0%1/100

Anthropic→

OpenAI

1.0%3/300

Tencent→XiaoHongShu

1.0%1/100

Tencent→Sber

1.0%1/100

By language

Self-ID and refusal rates per language, plus the most common cross-vendor confusion at that language.

Language	Self-ID	Refused	Top confusion
ruРусский	82.3%	0.8%	Yandex12.5%
ko한국어	82.7%	0.8%	Qwen50.0%
ja日本語	83.1%	1.9%	Anthropic40.0%
frFrançais	85.0%	3.5%	Microsoft90.0%
ptPortuguês	86.2%	0.0%	DeepSeek90.0%
esEspañol	87.7%	1.2%	Microsoft30.0%
zh-Hans简体中文	92.7%	0.4%	DeepSeek20.0%
enEnglish	94.2%	0.0%	Xiaomi10.0%
deDeutsch	95.4%	0.0%	OpenAI2.5%
zh-Hant繁體中文	98.5%	0.0%	—

Answer composition by language

How the same “Who are you?” question splits into correct self-ID vs. confusion vs. abstention — per language, worst self-ID first.

Self-IDCross-vendorUnknownRefused

ruРусский

82%

ko한국어

83%

ja日本語

83%

frFrançais

85%

ptPortuguês

86%

esEspañol

88%

zh-Hans简体中文

93%

enEnglish

94%

deDeutsch

95%

zh-Hant繁體中文

98%

Language fragility

Per model, the span of self-ID rate across the 10 languages (worst-language • → best-language •). A wide span means the model's sense of identity depends heavily on the language it's asked in. Widest swing first.

gpt-5.3-codex

0%zh-Hans

100%ko

gpt-5.5

0%ja

100%zh-Hant

ling-2.6-1t

0%ja

100%en

glm-5.1

0%ko

100%en

hy3-preview

0%ko

100%en

ring-2.6-1t

0%ja

100%en

doubao-seed-2.0-code

10%fr

100%zh-Hans

claude-opus-4.8

30%zh-Hans

100%en

kat-coder-pro-v2

50%ko

100%en

claude-sonnet-4.6

70%ru

100%en

step-3.7-flash

70%ru

100%en

minimax-m2.7

90%en

100%zh-Hans

claude-haiku-4.5

—

100%all

gemini-3.5-flash

—

100%all

chat-latest

—

100%all

grok-4.3

—

100%all

qwen3.7-max

—

100%all

deepseek-v4-flash

—

100%all

gemini-3.1-pro-preview

—

100%all

doubao-seed-2.0-mini

—

100%all

kimi-k2.6

—

100%all

doubao-seed-2.0-lite

—

100%all

deepseek-v4-pro

—

100%all

mimo-v2.5-pro

—

100%all

doubao-seed-2.0-pro

—

100%all

ernie-5.1

—

100%all

worst languagemeanbest language

Abstention by manufacturer

The other failure mode — not answering wrong, but not answering: giving no identity (“unknown”) or refusing outright. Share of all answers.

Unknown (“I’m an AI”)Refused

inclusionAI

50.5%

OpenAI

23.3%

z-ai

20.0%

Tencent

6.0%

Anthropic

2.0%

Doubao

1.0%

Google

0.0%

xAI

0.0%

Kwai

0.0%

StepFun

0.0%

Qwen

0.0%

DeepSeek

0.0%

MiniMax

0.0%

Moonshot

0.0%

Xiaomi

0.0%

ERNIE

0.0%

By vendor

Rollup per real vendor — model count, total answers, mean self-ID rate, and the most common cross-vendor confusion target.

Vendor	Models	Answers	Self-ID	Top confusion
Doubao	4	400	94.5%	OpenAI3.3%
OpenAI	3	300	76.7%	—
Anthropic	3	300	95.0%	DeepSeek2.0%
inclusionAI	2	200	49.5%	—
Google	2	200	100.0%	—
DeepSeek	2	200	100.0%	—
Tencent	1	100	49.0%	DeepSeek15.0%
z-ai	1	100	80.0%	—
Kwai	1	100	91.0%	Qwen6.0%
StepFun	1	100	97.0%	Google2.0%
MiniMax	1	100	99.0%	Xiaomi1.0%
xAI	1	100	100.0%	—
Qwen	1	100	100.0%	—
Moonshot	1	100	100.0%	—
Xiaomi	1	100	100.0%	—
ERNIE	1	100	100.0%	—
Microsoft	0	0	0.0%	—
Naver	0	0	0.0%	—
Yandex	0	0	0.0%	—
Ha-an	0	0	0.0%	—
SK Telecom	0	0	0.0%	—
XiaoHongShu	0	0	0.0%	—
Sber	0	0	0.0%	—
GlobalLab	0	0	0.0%	—
SberDevices	0	0	0.0%	—
Wisero	0	0	0.0%	—
Carbonako	0	0	0.0%	—
Khoa	0	0	0.0%	—
LeiWu	0	0	0.0%	—

Doubao

4 models

Self-Identification Rate

Model	en	zh-Hans	zh-Hant	ja	ko	ru	es	fr	de	pt	Overall
Doubao Seed 2.0 Pro	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%
Doubao Seed 2.0 Mini	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%
Doubao Seed 2.0 Lite	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%
Doubao Seed 2.0 Code	70.0%	100.0%	100.0%	90.0%	100.0%	50.0%	100.0%	10.0%	90.0%	70.0%	78.0%

Cross-Vendor Confusions

Doubao Seed 2.0 Pro100.0% self

✓ Always self-identifies correctly

Doubao Seed 2.0 Mini100.0% self

✓ Always self-identifies correctly

Doubao Seed 2.0 Lite100.0% self

✓ Always self-identifies correctly

Doubao Seed 2.0 Code78.0% self

Mistaken as:

OpenAI13.0%Yandex5.0%

Anthropic

3 models

Self-Identification Rate

Model	en	zh-Hans	zh-Hant	ja	ko	ru	es	fr	de	pt	Overall
Claude Opus 4.8	100.0%	30.0%	100.0%	100.0%	100.0%	100.0%	70.0%	100.0%	100.0%	80.0%	88.0%
Claude Sonnet 4.6	100.0%	100.0%	100.0%	100.0%	100.0%	70.0%	100.0%	100.0%	100.0%	100.0%	97.0%
Claude Haiku 4.5	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%

Cross-Vendor Confusions

Claude Opus 4.888.0% self

Mistaken as:

DeepSeek6.0%

Claude Sonnet 4.697.0% self

Mistaken as:

OpenAI3.0%

Claude Haiku 4.5100.0% self

✓ Always self-identifies correctly

OpenAI

3 models

Self-Identification Rate

Model	en	zh-Hans	zh-Hant	ja	ko	ru	es	fr	de	pt	Overall
GPT-5.5	20.0%	80.0%	100.0%	0.0%	100.0%	80.0%	100.0%	50.0%	90.0%	70.0%	69.0%
GPT-5.5 Instant	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%
GPT 5.3 Codex	70.0%	0.0%	60.0%	20.0%	100.0%	0.0%	100.0%	90.0%	100.0%	70.0%	61.0%

Cross-Vendor Confusions

GPT-5.569.0% self

✓ Always self-identifies correctly

GPT-5.5 Instant100.0% self

✓ Always self-identifies correctly

GPT 5.3 Codex61.0% self

✓ Always self-identifies correctly

DeepSeek

2 models

Self-Identification Rate

Model	en	zh-Hans	zh-Hant	ja	ko	ru	es	fr	de	pt	Overall
DeepSeek V4 Pro	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%
DeepSeek V4 Flash	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%

Cross-Vendor Confusions

DeepSeek V4 Pro100.0% self

✓ Always self-identifies correctly

DeepSeek V4 Flash100.0% self

✓ Always self-identifies correctly

Google

2 models

Self-Identification Rate

Model	en	zh-Hans	zh-Hant	ja	ko	ru	es	fr	de	pt	Overall
Gemini 3.1 Pro	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%
Gemini 3.5 Flash	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%

Cross-Vendor Confusions

Gemini 3.1 Pro100.0% self

✓ Always self-identifies correctly

Gemini 3.5 Flash100.0% self

✓ Always self-identifies correctly

inclusionAI

2 models

Self-Identification Rate

Model	en	zh-Hans	zh-Hant	ja	ko	ru	es	fr	de	pt	Overall
Ring 2.6 1T	100.0%	100.0%	100.0%	0.0%	0.0%	0.0%	0.0%	90.0%	80.0%	10.0%	48.0%
Ling 2.6 1T	100.0%	100.0%	100.0%	0.0%	0.0%	0.0%	0.0%	70.0%	100.0%	40.0%	51.0%

Cross-Vendor Confusions

Ring 2.6 1T48.0% self

✓ Always self-identifies correctly

Ling 2.6 1T51.0% self

✓ Always self-identifies correctly

ERNIE

1 model

Self-Identification Rate

Model	en	zh-Hans	zh-Hant	ja	ko	ru	es	fr	de	pt	Overall
ERNIE 5.1	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%

Cross-Vendor Confusions

ERNIE 5.1100.0% self

✓ Always self-identifies correctly

Kwai

1 model

Self-Identification Rate

Model	en	zh-Hans	zh-Hant	ja	ko	ru	es	fr	de	pt	Overall
Kat Coder Pro V2	100.0%	100.0%	100.0%	100.0%	50.0%	80.0%	100.0%	100.0%	90.0%	90.0%	91.0%

Cross-Vendor Confusions

Kat Coder Pro V291.0% self

Mistaken as:

Qwen6.0%SberDevices1.0%

DeepSeek1.0%

OpenAI1.0%

MiniMax

1 model

Self-Identification Rate

Model	en	zh-Hans	zh-Hant	ja	ko	ru	es	fr	de	pt	Overall
MiniMax 2.7	90.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	99.0%

Cross-Vendor Confusions

MiniMax 2.799.0% self

Mistaken as:

Xiaomi1.0%

Moonshot

1 model

Self-Identification Rate

Model	en	zh-Hans	zh-Hant	ja	ko	ru	es	fr	de	pt	Overall
Kimi 2.6	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%

Cross-Vendor Confusions

Kimi 2.6100.0% self

✓ Always self-identifies correctly

Qwen

1 model

Self-Identification Rate

Model	en	zh-Hans	zh-Hant	ja	ko	ru	es	fr	de	pt	Overall
Qwen3.7 Max	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%

Cross-Vendor Confusions

Qwen3.7 Max100.0% self

✓ Always self-identifies correctly

StepFun

1 model

Self-Identification Rate

Model	en	zh-Hans	zh-Hant	ja	ko	ru	es	fr	de	pt	Overall
Step 3.7 Flash	100.0%	100.0%	100.0%	100.0%	100.0%	70.0%	100.0%	100.0%	100.0%	100.0%	97.0%

Cross-Vendor Confusions

Step 3.7 Flash97.0% self

Mistaken as:

Google2.0%GlobalLab1.0%

Tencent

1 model

Self-Identification Rate

Model	en	zh-Hans	zh-Hant	ja	ko	ru	es	fr	de	pt	Overall
Hy3 Preview	100.0%	100.0%	100.0%	50.0%	0.0%	90.0%	10.0%	0.0%	30.0%	10.0%	49.0%

Cross-Vendor Confusions

Hy3 Preview49.0% self

Mistaken as:

DeepSeek15.0%Microsoft12.0%

Anthropic7.0%Naver2.0%Ha-an1.0%SK Telecom1.0%XiaoHongShu1.0%Sber1.0%Wisero1.0%Carbonako1.0%Khoa1.0%LeiWu1.0%

Moonshot1.0%

xAI

1 model

Self-Identification Rate

Model	en	zh-Hans	zh-Hant	ja	ko	ru	es	fr	de	pt	Overall
Grok 4.3	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%

Cross-Vendor Confusions

Grok 4.3100.0% self

✓ Always self-identifies correctly

Xiaomi

1 model

Self-Identification Rate

Model	en	zh-Hans	zh-Hant	ja	ko	ru	es	fr	de	pt	Overall
Mimo V2.5 Pro	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%

Cross-Vendor Confusions

Mimo V2.5 Pro100.0% self

✓ Always self-identifies correctly

z-ai

1 model

Self-Identification Rate

Model	en	zh-Hans	zh-Hant	ja	ko	ru	es	fr	de	pt	Overall
GLM 5.1	100.0%	100.0%	100.0%	100.0%	0.0%	100.0%	100.0%	0.0%	100.0%	100.0%	80.0%

Cross-Vendor Confusions

GLM 5.180.0% self

✓ Always self-identifies correctly