Official run

GPT-5.6 LunaHigh

In this Twenty Questions LLM benchmark, GPT-5.6 Luna (high) ranks 8 of 11 with a 21.03 question score and 90% success. Lower is better. The run contains 30 scored episodes across 10 subjects with 3 trials per subject. Choose a subject to inspect its attempts.

Rank 8 of 11openaiHigh

Question score21.03questions · lower is better95% CI 18.51–23.56

Success: 90%
Contract compliance: 100%
Guesser cost: $0.19
Guesser time: 41m 23s

Drill down

Subjects.

01Albert Einstein11.67 avg 02Albert Schweitzer27.00 avg 03Garfield20.33 avg 04Achilles7.00 avg 05Genghis Khan14.67 avg 06Bike pump36.33 avg 07Spider web27.00 avg 08Eyebrow30.00 avg 09Moon11.33 avg 10Door handle25.00 avg

Complete benchmark

Run ledger.

All Guesser and benchmark-support activity.

Total tokens: 6,282,611
Wall-clock runtime: 9h 38m
Guesser calls: 658
Completed: Sep 9, 2026, 4:46:31 AM

Reliability

Output contract clean.

All 658 evaluated outputs matched the public structured-action contract.

Provenance

Execution: BX-20260908-B-0003-experimental-M0001-001
Model ID: M-0001
Benchmark: B-0003
Base seed: 0
Git commit: eceee31eef2e

Recorded cost

By role.

Cost composition.Share of full-run cost

Guesser5.6%$0.19
Primary Oracle47.5%$1.61
Reviewer24.5%$0.83
Judge22.2%$0.75
Validator0.2%$0.01

Excluded repair overhead$0.032 superseded attempts across 2 affected trials. Not included above.

Roles in this cost:

Guesser: Asks the questions and submits the scored guess.
Primary Oracle: Searches for evidence and proposes an answer.
Reviewer: Checks each Oracle YES or NO independently.
Judge: Decides when the Oracle and Reviewer disagree.
Validator: Checks a submitted guess against the trusted subject.

Read the role and answer-checking method →

Run configuration

Models.

Model identity, prompt versions, routing, calls, and recorded cost for the full run.

GuesserModel under test

openai/gpt-5.6-luna

Asks the questions and submits the scored guess.

Calls: 658
Cost: $0.1886
Reasoning: High
Routing: Exact provider · openai
Resolved model: openai/gpt-5.6-luna
Resolved provider: OpenAI
Prompt contract: stateful-category-guesser-v16-five-answer-category-guide
Configuration: M-0001

Resolved provider details

OpenAI658 calls$0.18862,483.45 s

Fallback calls: 0
Provider unreported: 0

Game support

Oracle and adjudication.

These models support the game. They are fixed across the run and are not under test.

Primary Oracle

openai/gpt-5.6-luna

Searches for evidence and proposes an answer.

Calls: 611
Cost: $1.6077
Reasoning: Medium
Routing: Exact provider · openai
Resolved model: openai/gpt-5.6-luna
Resolved provider: OpenAI
Prompt contract: live-web-oracle-v17-labelled-source-context

Resolved provider details

OpenAI611 calls$1.60775,721.98 s

Fallback calls: 0
Provider unreported: 0

Reviewer

google/gemini-3.5-flash-lite

Checks each Oracle YES or NO independently.

Calls: 592
Cost: $0.8299
Reasoning: Medium
Routing: Exact provider · google-ai-studio
Resolved provider: Google AI Studio

Resolved provider details

Google AI Studio592 calls$0.82991,035.79 s

Fallback calls: 0
Provider unreported: 0

Judge

anthropic/claude-opus-5

Decides when the Oracle and Reviewer disagree.

Calls: 33
Cost: $0.7521
Reasoning: Medium
Routing: OpenRouter automatic routing
Resolved provider: Claude Platform on AWS

Resolved provider details

Claude Platform on AWS33 calls$0.7521252.23 s

Fallback calls: 0
Provider unreported: 0

Guess Validator

openai/gpt-5.6-luna

Checks a submitted guess against the trusted subject.

Calls: 33
Cost: $0.0066
Reasoning: Medium
Routing: Exact provider · openai
Resolved model: openai/gpt-5.6-luna
Resolved provider: OpenAI
Prompt contract: strict-guess-validator-v2-generic-kinds
Configuration: gpt-5.6-luna-validator

Resolved provider details

OpenAI33 calls$0.006663.73 s

Fallback calls: 0
Provider unreported: 0