MICRO MOUSE BENCHMARK · LOCAL, THINKING LOW

gpt-oss 20B

Openai · open weights · released 2025-08-05 · route gpt-oss:20b · Model profile · Chess results · Sudoku results · Mines results · FreeCell results · Nations Royale results · Hustle Sim results · Elemental Armory results

Results

Score
095%: 0 to 28 · 5 of 14 base runs so far
Campaign score against the bot
00 to 14
Daily mazes faster than the median person
0 of 1
Reached the dock
0 of 5 mazes
Faster than the bot
0 of 5
Reached / out of time / out of replies
0 / 4 / 1
Share of the route covered, on average
9%
Wall hits per run
116.6
Drift boosts / pad boosts per run
0.0 / 0.0
Replies per run
82.8
Control segments per reply
1.75
Values moved into range
101
Segments cut at the one-second limit
200
Lines that were not segments
2
Replies with no readable segment
0.0%
Retries for an unreadable reply
5
Model minutes per run
613.4
Tokens per reply (in / out)
1,597 / 5,353
Reasoning tokens per reply
n/a
API-equivalent cost per run
n/a
Seconds per reply
444.5
Replies cut off at the length limit
0.7%
Aborted on our side (not counted)
4

By tier

TierAgainstRunsScoreReachedPaceWall hits / runTest pilot
Daily mazesmedian person100 of 1n/a103.048
Workbenchbot100 of 1n/a91.052
Cold Storagebot100 of 1n/a98.049
Fast Lanebot100 of 1n/a103.050
Night Shiftbot100 of 1n/a188.050

Every run