CHESS BENCHMARK · LOCAL, THINKING LOW

Qwen3.8 Flash-Next

Alibaba · open weights · released 2026-08-26 · route qwen3.8-flash-next · Model profile

Opening gap: -441 (95%: -1,107 to +225). The interval includes zero, so we can't yet say whether it performs differently across the variants.

Standard

Rating
20495%: -300 to 709
Record
0 won, 0 drawn, 4 lost of 4
Accuracy (open positions)
72.2
Accuracy (all moves)
78.2
Average centipawn loss
153
Blunders / mistakes / inaccuracies per 100 moves
17.8 / 5.9 / 9.9
Matched Stockfish's first choice
26%
In Stockfish's top three
50%
Legal on the first try
85.4%
Illegal / invalid
22 of 123 answers
Forfeits
2
Adjudicated games
1
API-equivalent cost per game
n/a
API-equivalent cost per move
n/a
Seconds per move
295.1
Model minutes per game
126.6
Tokens per move (in / out)
446 / 4,548
Reasoning tokens per move
n/a
Aborted on our side (not counted)
0

By phase

PhaseMovesAccuracyCentipawn lossBlunders /100Engine match
Opening4085.89810.048%
Middlegame3363.718227.312%
Endgame2848.719817.911%

Chess960

Rating
64595%: 210 to 1,080
Record
1 won, 0 drawn, 3 lost of 4
Accuracy (open positions)
72.4
Accuracy (all moves)
80.3
Average centipawn loss
156
Blunders / mistakes / inaccuracies per 100 moves
14.7 / 3.9 / 13.7
Matched Stockfish's first choice
24%
In Stockfish's top three
47%
Legal on the first try
77.5%
Illegal / invalid
24 of 126 answers
Forfeits
0
Adjudicated games
2
API-equivalent cost per game
n/a
API-equivalent cost per move
n/a
Seconds per move
292.0
Model minutes per game
124.1
Tokens per move (in / out)
544 / 5,273
Reasoning tokens per move
n/a
Aborted on our side (not counted)
0

By phase

PhaseMovesAccuracyCentipawn lossBlunders /100Engine match
Opening3674.312219.419%
Middlegame3158.824116.119%
Endgame3572.01168.631%

Games