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: -47 (95%: -560 to +466). The interval includes zero, so we can't yet say whether it performs differently across the variants.

Standard

Rating
41195%: 74 to 748
Record
2 won, 0 drawn, 4 lost of 6
Accuracy (open positions)
73.4
Accuracy (all moves)
79.8
Average centipawn loss
134
Blunders / mistakes / inaccuracies per 100 moves
15.9 / 4.8 / 10.3
Matched Stockfish's first choice
27%
In Stockfish's top three
49%
Legal on the first try
83.7%
Illegal / invalid
32 of 177 answers
Forfeits
2
Adjudicated games
3
API-equivalent cost per game
n/a
API-equivalent cost per move
n/a
Seconds per move
260.7
Model minutes per game
106.5
Tokens per move (in / out)
440 / 4,237
Reasoning tokens per move
n/a
Aborted on our side (not counted)
323

By phase

PhaseMovesAccuracyCentipawn lossBlunders /100Engine match
Opening6083.79911.742%
Middlegame5563.414420.018%
Endgame3048.718516.713%

Chess960

Rating
45895%: 71 to 844
Record
1 won, 0 drawn, 5 lost of 6
Accuracy (open positions)
71.1
Accuracy (all moves)
80.0
Average centipawn loss
154
Blunders / mistakes / inaccuracies per 100 moves
14.5 / 5.8 / 13.8
Matched Stockfish's first choice
24%
In Stockfish's top three
45%
Legal on the first try
76.3%
Illegal / invalid
36 of 174 answers
Forfeits
1
Adjudicated games
3
API-equivalent cost per game
n/a
API-equivalent cost per move
n/a
Seconds per move
310.2
Model minutes per game
119.8
Tokens per move (in / out)
542 / 5,349
Reasoning tokens per move
n/a
Aborted on our side (not counted)
1

By phase

PhaseMovesAccuracyCentipawn lossBlunders /100Engine match
Opening5273.412021.121%
Middlegame5158.521411.822%
Endgame3572.01168.631%

Games