ELEMENTAL ARMORY BENCHMARK · LOCAL, THINKING LOW

Qwen3.8 Flash-Next

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

Results

Score
6795%: 36 to 98 · 3 of 150 base matches so far
Matches won
2 of 3
Rounds won
5 of 9 (56%)
Drawn rounds, replayed
0
Rounds its ball countered the rival's
67%
Rounds the rival's ball countered its
0%
Average card rating
3.00
Card slots filled
100%
Decks it brought most
Fire/Dark/Plant/Ice (1); Fire/Plant/Wind/Ice (1); Fire/Water/Lightning/Light (1)
Replies the game could not accept
0.0%
Replies with no readable answer
0.0%
Decisions that fell back to the default
0.0%
Decks that fell back to the default
0
Requests per match
4.0
Tokens per decision (in / out)
1,811 / 3,314
Reasoning tokens per decision
n/a
Model seconds per match
775.5
API-equivalent cost per match
n/a
Seconds per decision
193.9
Replies cut off at the length limit
0.0%
Aborted on our side (not counted)
0

By tier

LevelMatchesWonRoundsStarsRival in its seatTuned player
Level 111 (100%)67%3.01 of 11 of 1
Level 510 (0%)33%3.00 of 11 of 1
Level 811 (100%)67%3.00 of 10 of 1

Against the game's players, match by match

On the same matches: the share it won that the player lost, minus the share the player won that it lost, in points. Better and worse count those matches.

PlayerMatchesGapBetter / worse
Rival in your seat3+33.31 / 0
Starter player3+33.31 / 0
Sharp player3+0.01 / 1
Tuned player3+0.01 / 1

Every match