Official run

Grok 4.6High

In this Twenty Questions LLM benchmark, Grok 4.6 (high) ranks 5 of 11 with a 17.13 question score and 100% 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 5 of 11xaiHigh

Question score17.13questions · lower is better95% CI 15.54–18.73

Success: 100%
Contract compliance: 100%
Guesser cost: $4.10
Guesser time: 3h 25m

Drill down

Subjects.

01Albert Einstein10.33 avg 02Albert Schweitzer30.33 avg 03Garfield14.00 avg 04Achilles8.00 avg 05Genghis Khan13.67 avg 06Bike pump27.00 avg 07Spider web22.00 avg 08Eyebrow16.00 avg 09Moon8.33 avg 10Door handle21.67 avg

Complete benchmark

Run ledger.

All Guesser and benchmark-support activity.

Total tokens: 5,758,939
Wall-clock runtime: 7h 2m
Guesser calls: 544
Completed: Sep 8, 2026, 6:36:46 PM

Reliability

Output contract clean.

All 544 evaluated outputs matched the public structured-action contract.

Provenance

Execution: BX-20260908-B-0003-experimental-M0015-001
Model ID: M-0015
Benchmark: B-0003
Base seed: 0
Git commit: eceee31eef2e

Recorded cost

By role.

Cost composition.Share of full-run cost

Guesser62.5%$4.10
Primary Oracle20.4%$1.34
Reviewer9.8%$0.65
Judge7.2%$0.47
Validator0.1%$0.01

Excluded repair overhead$0.534 superseded attempts across 3 affected trials.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

x-ai/grok-4.6

Asks the questions and submits the scored guess.

Calls: 544
Cost: $4.1005
Reasoning: High
Routing: Exact provider · xai
Resolved model: x-ai/grok-4.6
Resolved provider: xAI
Prompt contract: stateful-category-guesser-v16-five-answer-category-guide
Configuration: M-0015

Resolved provider details

xAI544 calls$4.100512,325.99 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: 512
Cost: $1.3396
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

OpenAI512 calls$1.33964,996.92 s

Fallback calls: 0
Provider unreported: 0

Reviewer

google/gemini-3.5-flash-lite

Checks each Oracle YES or NO independently.

Calls: 494
Cost: $0.6461
Reasoning: Medium
Routing: Exact provider · google-ai-studio
Resolved provider: Google AI Studio

Resolved provider details

Google AI Studio494 calls$0.6461850.68 s

Fallback calls: 0
Provider unreported: 0

Judge

anthropic/claude-opus-5

Decides when the Oracle and Reviewer disagree.

Calls: 21
Cost: $0.4697
Reasoning: Medium
Routing: OpenRouter automatic routing
Resolved provider: Claude Platform on AWS

Resolved provider details

Claude Platform on AWS21 calls$0.4697163.09 s

Fallback calls: 0
Provider unreported: 0

Guess Validator

openai/gpt-5.6-luna

Checks a submitted guess against the trusted subject.

Calls: 32
Cost: $0.0064
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

OpenAI32 calls$0.006465.49 s

Fallback calls: 0
Provider unreported: 0