Model summary · Edition 1.1

gpt-oss-120BHigh

These results come from one run: 10 fixed subjects, 3 rounds each, a 40-question limit, and reasoning setting high. They do not predict results on other subjects or settings.

This page always shows the model’s newest ranked run.

See all 30 rounds → All models

Every question, answer, guess, and source.

Rank: Rank 18of 20 models
Question score: 22.23 Lower is better. 95% confidence interval 19.58 to 24.89.
Rounds solved: 20 of 30 71 format errors
Guesser cost per round: $0.0682 Model time 1m 5s.

See how gpt-oss-120B (high) played. All 30 rounds, with every question, answer, guess, and source.

Observations

What stood out in this run.

Counts and ranks from this run only. A subject rank compares this model’s average on that subject with every other model in this edition. Each item links to its rounds.

Best ranks vs other models

Albert Einstein: best of 20 models, average 4.67 questions. round 1, round 2, round 3
Achilles: best of 20 models, average 7.00 questions. round 1, round 2, round 3

Lowest ranks vs other models

Eyebrow: 20th of 20 models, average 41.00 questions. round 1, round 2, round 3
Spider web: 19th of 20 models, average 35.33 questions. round 1, round 2, round 3

Single rounds

Fewest questions in a solved round: 4, on Albert Einstein. round 2
Most questions in a solved round: 25, on Albert Schweitzer. round 2
Largest spread between rounds: Door handle, from 18 to 41 questions. round 1, round 2, round 3
18 wrong guesses in 30 rounds, up to 6 in one round on Eyebrow. round 1

Failed rounds and format errors

Not solved: Bike pump. Too many format errors after 40 questions; scored as 41. round 1
Not solved: Bike pump. Too many format errors after 40 questions; scored as 41. round 2
Not solved: Bike pump. Too many format errors after 40 questions; scored as 41. round 3
Not solved: Spider web. Too many format errors after 37 questions; scored as 41. round 1
Not solved: Spider web. Too many format errors after 40 questions; scored as 41. round 2
Not solved: Eyebrow. Too many format errors after 32 questions; scored as 41. round 1
Not solved: Eyebrow. Too many format errors after 40 questions; scored as 41. round 2
Not solved: Eyebrow. Too many format errors after 40 questions; scored as 41. round 3
Not solved: Door handle. Too many format errors after 40 questions; scored as 41. round 1
Not solved: Door handle. Too many format errors after 30 questions; scored as 41. round 2
71 format errors in 13 rounds. Each one before the question limit used a counted turn. Albert Schweitzer round 1, Albert Schweitzer round 2, Albert Schweitzer round 3, Bike pump round 1, Bike pump round 2, Bike pump round 3, Spider web round 1, Spider web round 2, Eyebrow round 1, Eyebrow round 2, Eyebrow round 3, Door handle round 1, Door handle round 2

Subjects

Results by subject.

Failed rounds count as 41 questions. Rank compares this subject's average with the other models in edition 1.1. Select a subject or a round to see the game.

Subject	Average questions	Subject rank	Solved	Rounds
Albert Einstein	4.67	1 of 20	3 of 3	6, 4, 4
Albert Schweitzer	22.67	11 of 20	3 of 3	18, 25, 25
Garfield	10.00	4 of 20	3 of 3	7, 7, 16
Achilles	7.00	1 of 20	3 of 3	7, 9, 5
Genghis Khan	16.33	18 of 20	3 of 3	15, 15, 19
Bike pump	41.00	17 of 20	0 of 3	not solved, not solved, not solved
Spider web	35.33	19 of 20	1 of 3	not solved, not solved, 24
Eyebrow	41.00	20 of 20	0 of 3	not solved, not solved, not solved
Moon	11.00	16 of 20	3 of 3	8, 12, 13
Door handle	33.33	17 of 20	1 of 3	not solved, not solved, 18

Runs

Runs of this model.

Each edition ranks only the newest complete run of a model. Earlier runs are not ranked and their details are not published.

Current run in edition 1.1: BX-20260908-B-0003-experimental-M0002-001, completed 8 Sep 2026.
Edition 1.0: rank 6 of 18, question score 13.63. Different rules and subjects, so scores are not comparable across editions.