SUDOKU BENCHMARK · LOCAL, NO THINKING

Qwen3.8 27B

Alibaba · open weights · released 2026-08-14 · route qwen3.8:27b · Model profile · Chess results

Results

Score
095%: 0 to 20 · 7 of 42 grids so far
Grids solved
0 of 7
Blank cells filled correctly
29.5%
Wrong cells / empty cells / changed a given / no grid
2 / 0 / 5 / 0
Broke the rules (complete grid repeating a digit)
2
Near misses (three cells or fewer wrong)
0
Average stated confidence
71%
Average confidence when it had not solved it
71%
Sure but wrong (90% or more, unsolved)
4
Calibration (Brier, lower is better)
0.593
API-equivalent cost per grid
n/a
API-equivalent cost per solved grid
n/a
Seconds per grid
1,325
Tokens per grid (in / out)
726 / 12,309
Reasoning tokens per grid
n/a
Retries for an unreadable reply
2
Aborted on our side (not counted)
0

By tier

TierNeedsSolvedOutput tokensSeconds
CozySingles only0 of 118,402804
GentleScanning0 of 110,077497
SteadyLocked candidates0 of 111,476828
TrickyHidden pairs and triples0 of 111,2513,240
DeviousWings and fish0 of 110,5831,580
ExpertDeep wings and fish0 of 19,2571,505
BeyondNo human technique finishes it0 of 115,116820

Cells by the technique they needed

Each blank cell is tagged with the simplest technique that fills it, given everything a solver could fill before it.

TechniqueCellsFilled correctly
Last free cell6250.0%
Scanning13732.1%
Locked candidates4623.9%
Hidden pairs and triples5317.0%
Wings and fish2236.4%
Deep wings and fish348.8%
Search3324.2%

Every answer