SUDOKU BENCHMARK · LOCAL, NO THINKING

Qwen3.5 9B

Alibaba · open weights · released 2026-02-23 · route qwen3.5:9b · Model profile · Chess results

Results

Score
095%: 0 to 16 · 15 of 42 grids so far
Grids solved
0 of 15
Blank cells filled correctly
8.5%
Wrong cells / empty cells / changed a given / no grid
0 / 1 / 7 / 7
Broke the rules (complete grid repeating a digit)
0
Near misses (three cells or fewer wrong)
0
Average stated confidence
0%
Average confidence when it had not solved it
0%
Sure but wrong (90% or more, unsolved)
0
Calibration (Brier, lower is better)
0.000
API-equivalent cost per grid
n/a
API-equivalent cost per solved grid
n/a
Seconds per grid
220
Tokens per grid (in / out)
1,620 / 2,342
Reasoning tokens per grid
n/a
Retries for an unreadable reply
14
Aborted on our side (not counted)
0

By tier

TierNeedsSolvedOutput tokensSeconds
CozySingles only0 of 32,071252
GentleScanning0 of 23,000248
SteadyLocked candidates0 of 23,000293
TrickyHidden pairs and triples0 of 23,000248
DeviousWings and fish0 of 21,553177
ExpertDeep wings and fish0 of 22,302114
BeyondNo human technique finishes it0 of 21,606188

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 cell10120.8%
Scanning14414.6%
Locked candidates156.7%
Hidden pairs and triples20.0%
Wings and fish618.2%
Deep wings and fish3429.4%
Search8113.6%

Every answer