Official run

Muse WonderHigh

In this Twenty Questions AI benchmark, Muse Wonder (high) ranks 14 of 20 with a question score of 19.63: the average number of questions per round, where lower is better. It failed 1 of 30 rounds. The run has 30 scored rounds: 10 subjects, 3 rounds each. Choose a subject to see its rounds.

Rank 14 of 20metaHigh

Question score19.63questions per roundLower is better · 95% CI 17.58–21.68

Solved: 97%
Valid format: 100%
Guesser cost: $4.82
Guesser time: 53m 29s

Drill down

Subjects.

01Albert Einstein14.33 avg 02Albert Schweitzer22.33 avg 03Garfield20.67 avg 04Achilles8.00 avg 05Genghis Khan15.33 avg 06Bike pump37.33 avg 07Spider web22.33 avg 08Eyebrow20.00 avg 09Moon9.33 avg 10Door handle26.67 avg

Complete benchmark

Run totals.

Includes the Guesser and all answer-checking models.

Total tokens: 4,449,285
Total runtime: 2h 24m
Guesser calls: 618
Completed: 13 Oct 2026, 08:20 UTC

Reply format

No format errors.

All 618 checked replies used the required format.

Run details

Execution: BX-20261013-B-0003-OFFICIAL-M0033-002
Model ID: M-0033
Benchmark: B-0003
Base seed: 0
Git commit: 333294db9cac

Recorded cost

By role.

Cost composition.Share of full-run cost

Guesser67.1%$4.82
Oracle15.3%$1.10
Reviewer7.8%$0.56
Judge9.8%$0.70
Validator0.1%$0.01

Rerun cost (excluded)$0.075 earlier attempts in 5 rerun rounds. Not included above.

Roles in this cost:

Guesser: The tested model. Asks questions and makes guesses.
Oracle: Searches for evidence and proposes an answer.
Reviewer: Checks each Oracle answer except Unknown, without seeing it.
Judge: Decides when the Oracle and Reviewer disagree.
Validator: Checks whether a guess names the hidden subject.

How the roles and answer checks work →

Run configuration

Models.

Models, prompt versions, providers, calls, and cost for the whole run.

GuesserModel under test

meta/muse-wonder

The tested model. Asks questions and makes guesses.

Calls: 618
Cost: $4.8162
Reasoning: High
Routing: Fixed provider · meta
Resolved model: meta/muse-wonder
Resolved provider: Meta
Prompt version: stateful-category-guesser-v16-five-answer-category-guide
Configuration: M-0033

Provider details

Meta618 calls$4.81623,208.78 s

Fallback calls: 0
Provider not reported: 0

Answer checks

Answering and checking models.

These models answer and check the Guesser's questions and guesses. They are the same for every round and are not being tested.

Oracle

openai/gpt-5.6-luna

Searches for evidence and proposes an answer.

Calls: 436
Cost: $1.0982
Reasoning: Medium
Routing: Fixed provider · openai
Resolved model: openai/gpt-5.6-luna
Resolved provider: OpenAI
Prompt version: live-web-oracle-v17-labelled-source-context

Provider details

OpenAI436 calls$1.09823,810.42 s

Fallback calls: 0
Provider not reported: 0

Reviewer

google/gemini-3.5-flash-lite

Checks each Oracle answer except Unknown, without seeing it.

Calls: 412
Cost: $0.5581
Reasoning: Medium
Routing: Fixed provider · google-ai-studio
Resolved provider: Google AI Studio

Provider details

Google AI Studio412 calls$0.5581699.08 s

Fallback calls: 0
Provider not reported: 0

Judge

anthropic/claude-opus-5

Decides when the Oracle and Reviewer disagree.

Calls: 30
Cost: $0.7037
Reasoning: Medium
Routing: Chosen automatically by OpenRouter
Resolved provider: Claude Platform on AWS

Provider details

Claude Platform on AWS30 calls$0.7037188.65 s

Fallback calls: 0
Provider not reported: 0

Guess Validator

openai/gpt-5.6-luna

Checks whether a guess names the hidden subject.

Calls: 30
Cost: $0.0056
Reasoning: Medium
Routing: Fixed provider · openai
Resolved model: openai/gpt-5.6-luna
Resolved provider: OpenAI
Prompt version: strict-guess-validator-v2-generic-kinds
Configuration: gpt-5.6-luna-validator

Provider details

OpenAI30 calls$0.005645.99 s

Fallback calls: 0
Provider not reported: 0