CHESS BENCHMARK · LOCAL, NO THINKING

Qwen3.5 9B

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

Opening gap: +0 (95%: -813 to +813). The interval includes zero, so we can't yet say whether it performs differently across the variants.

Standard

Rating
33295%: -243 to 907
Record
0 won, 0 drawn, 3 lost of 3
Accuracy (open positions)
81.6
Accuracy (all moves)
81.6
Average centipawn loss
57
Blunders / mistakes / inaccuracies per 100 moves
0.0 / 33.3 / 0.0
Matched Stockfish's first choice
0%
In Stockfish's top three
44%
Legal on the first try
50.0%
Illegal / invalid
12 of 21 answers
Forfeits
3
Adjudicated games
0
API-equivalent cost per game
n/a
API-equivalent cost per move
n/a
Seconds per move
98.2
Model minutes per game
6.5
Tokens per move (in / out)
1,116 / 1,158
Reasoning tokens per move
n/a
Aborted on our side (not counted)
0

By phase

PhaseMovesAccuracyCentipawn lossBlunders /100Engine match
Opening981.6570.00%
Middlegame0n/an/an/an/a
Endgame0n/an/an/an/a

Chess960

Rating
33295%: -243 to 907
Record
0 won, 0 drawn, 3 lost of 3
Accuracy (open positions)
83.5
Accuracy (all moves)
83.5
Average centipawn loss
49
Blunders / mistakes / inaccuracies per 100 moves
0.0 / 33.3 / 0.0
Matched Stockfish's first choice
0%
In Stockfish's top three
0%
Legal on the first try
0.0%
Illegal / invalid
13 of 16 answers
Forfeits
3
Adjudicated games
0
API-equivalent cost per game
n/a
API-equivalent cost per move
n/a
Seconds per move
264.2
Model minutes per game
8.8
Tokens per move (in / out)
1,894 / 2,260
Reasoning tokens per move
n/a
Aborted on our side (not counted)
0

By phase

PhaseMovesAccuracyCentipawn lossBlunders /100Engine match
Opening383.5490.00%
Middlegame0n/an/an/an/a
Endgame0n/an/an/an/a

Games