Official run

Kimi K3High

In this Twenty Questions LLM benchmark, Kimi K3 (high) ranks 4 of 18 with a 12.74 question score and 94% success. Lower is better. The run contains 35 scored episodes across 7 subjects with 5 trials per subject. Choose a subject to inspect its attempts.

Rank 4 of 18moonshotaiHigh

Question score12.74questions · lower is better95% CI 9.16–16.33

Success: 94%
Contract compliance: 98%
Guesser cost: $7.11
Guesser time: 4h 2m

Drill down

Subjects.

01Albert Einstein7.60 avg 02Albert Schweitzer27.20 avg 03Stephen King13.80 avg 04Garfield17.80 avg 05Achilles8.20 avg 06Genghis Khan10.40 avg 07Mario4.20 avg

Complete benchmark

Run ledger.

All Guesser and benchmark-support activity.

Total tokens: 4,323,541
Wall-clock runtime: 5h 17m
Guesser calls: 479
Completed: Jul 28, 2026, 10:10:44 PM

Reliability

Output contract breached.

11 invalid outputs affected 10 episodes and consumed 11 counted turns.

Provenance

Execution: BX-20260728-official-M0007-010
Model ID: M-0007
Benchmark: B-0001
Base seed: 0
Git commit: 4c694d834a90

Recorded cost

By role.

Cost composition.Share of full-run cost

Guesser70.5%$7.11
Primary Oracle22%$2.22
Reviewer3.6%$0.37
Judge3.7%$0.37
Validator0.2%$0.02

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

moonshotai/kimi-k3

Asks the questions and submits the scored guess.

Calls: 479
Cost: $7.1126
Reasoning: High
Routing: Exact provider · moonshotai
Resolved model: moonshotai/kimi-k3
Resolved provider: Moonshot AI
Prompt contract: stateful-category-guesser-v10-unknown-evidence-guidance
Configuration: M-0007

Resolved provider details

Legacy run. Per-call routing totals were not retained. The configured provider was moonshotai.

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: 415
Cost: $2.2205
Reasoning: Medium
Routing: Exact provider · openai
Resolved model: openai/gpt-5.6-luna
Resolved provider: OpenAI
Prompt contract: live-web-oracle-v7-direct-negative-evidence

Resolved provider details

Legacy run. Per-call routing totals were not retained. The configured provider was openai.

Reviewer

google/gemini-3.5-flash-lite

Checks each Oracle YES or NO independently.

Calls: 378
Cost: $0.3666
Reasoning: Medium
Routing: Exact provider · google-ai-studio

Resolved provider details

Legacy run. Per-call routing totals were not retained. The configured provider was google-ai-studio.

Judge

anthropic/claude-opus-5

Decides when the Oracle and Reviewer disagree.

Calls: 31
Cost: $0.3705
Reasoning: Medium
Routing: Exact provider · anthropic

Resolved provider details

Legacy run. Per-call routing totals were not retained. The configured provider was anthropic.

Guess Validator

openai/gpt-5.6-luna

Checks a submitted guess against the trusted subject.

Calls: 51
Cost: $0.0165
Reasoning: Medium
Routing: Exact provider · openai
Resolved model: openai/gpt-5.6-luna
Resolved provider: OpenAI
Prompt contract: strict-guess-validator-v1
Configuration: gpt-5.6-luna-validator

Resolved provider details

Legacy run. Per-call routing totals were not retained. The configured provider was openai.