Who's that LLM?

Statistics

Solve rate
55.6%

5 of 9 games

Failed
33.3%

3 of 9 games

Give-ups
11.1%

1 of 9 games

Correct guesses
45.5%

5 of 11 guesses

Median messages
2.0

Per game played

Median tokens
2,679

Prompt plus completion

Games over time

Games started per UTC day, last 14 days.

Which models are hardest to find

Hidden models, hardest first. Solve rate is solved over games played. Rounds given up before any message are left out.

Hidden modelGamesSolve rateMessages when solvedTokens
Llama 4 Scout10%n/a9,359
Mistral Small 310%n/a11.6K
Gemini 3.1 Pro2100%6.018.6K
DeepSeek V4 Pro1100%2.02,335
GPT-4o mini1100%2.01,023
Qwen3.8 Max1100%3.02,679

Models people mix up

Wrong names on games. Each row is the hidden model and the one that was named instead.

Hidden modelNamed insteadTimes
DeepSeek V4 ProGPT-6.1 Sol1
Gemini 3.1 ProClaude Sonnet 5.51
Llama 4 ScoutQwen3.8 Max1
Mistral Small 3Mixtral 8x22B1
Qwen3.8 MaxDeepSeek V4 Pro1
Qwen3.8 MaxGemini 3.1 Pro1

Size

By the size of the hidden model.

SizeGamesSolve rateMessages when solvedTokens per gameAverage score
Small425.0%2.05,5075,510
Large4100%4.35,9056,449

Origin

By the origin of the hidden model. Origin found is a game where a guess got the origin.

OriginGamesSolve rateMessages when solvedOrigin foundTokens per gameAverage score
USA560.0%4.760.0%5,7986,228
China2100%2.5100%2,5078,707
Europe10%n/a100%11.6K0

Difficulty

By the difficulty the player chose. Only Daily and All models have one.

DifficultyGamesSolve rateMessages when solvedTokens per gameAverage score
Easy5100%3.84,9298,465
Medium10%n/a9,3590
Hard10%n/a0n/a
Impossible0n/an/an/an/a

Hints

Rounds where hints could be bought against rounds played with no hints.

HintsGamesSolve rateMessages when solvedTokens per gameAverage score
With hints862.5%3.85,7066,047
No hints0n/an/an/an/a