DuelLabGameBench 2
Menu

Model results · Moonshot

Kimi K3

Rank 39 of 66 in Overall Models. All three benchmark settings tested.

Score · relative 0–100
48.8
Rank
39
Status
Complete 3/3
Rank change
-1
Players generated
2026-07-24 – 2026-10-09
Match evidence updated
2026-10-26
Results published
Compare in charts View player programs Back to leaderboard

Scores by reasoning setting

  • Max (Intensive group, tested and included)45.1
  • High (Balanced group, tested and included)49.6
  • Low (Baseline group, tested and included)51.6

Every tested setting is included in the overall score.

Available settings and grouping

Available settings

LowHighMax

Checked on 2026-07-24 · Provider documentation

How settings are grouped

  • Baseline: Low — tested and included
  • Balanced: High — tested and included
  • Intensive: Max — tested and included

Scores by game

  • Go42.5
  • Chess42.1
  • Shogi64.5
  • Xiangqi48.8
  • Hex56.4
  • International Draughts1 of 3 failed34.6
  • Game of the Amazons48.9
  • Tumbleweed52.4

Each game counts every generation attempt, a failed one as 0 (“1 of 2 failed” marks games with failed attempts). Scores are relative within each game, across the included settings; they do not compare game difficulty.

What this result shows

Strongest included setting: Low (51.6).

Across included settings, the highest game score is Shogi (64.5); the lowest is International Draughts (34.6). Each game counts every generation attempt, a failed one as 0. These are relative scores within each game, not direct comparisons of difficulty.

Match evidence measures these retained programs. It does not establish how a fresh generation will perform. Independent-generation variation is not estimated in this release.

Test protocol and evidence limits
Model identifier
moonshotai/kimi-k3
Provider route (catalog)
openrouter
Prompt identity / profile
Not retained in this result set / Not retained in this result set
Prompt version
Not retained in this result set
Generation policy
Generation-quality admission policy (see methodology)
Requested output-token limits
262144 (10 of 35 indexed attempts)

Generation evidence

  • Low: 8 contributing player programs across 8 games
  • High: 8 contributing player programs across 8 games
  • Max: 7 contributing player programs across 7 games

Counts describe contributing programs, not independent replications: repairs and resends are not new independent samples. Independent-generation lineage is not retained.

GameReasoning groupProgramsWorking-program strengthRaw ratingCombined record (W / D / L)
GoIntensive1 working of 1 attempt6.6891.726 / 8 / 482
GoBalanced1 working of 1 attempt69.11661.1301 / 89 / 126
GoBaseline1 working of 1 attempt51.71485.5125 / 262 / 129
ChessIntensive1 working of 1 attempt42.91417.029 / 338 / 113
ChessBalanced1 working of 1 attempt41.41405.515 / 352 / 113
ChessBaseline1 working of 1 attempt42.11410.47 / 374 / 99
ShogiIntensive1 working of 1 attempt77.81730.9354 / 32 / 108
ShogiBalanced1 working of 1 attempt66.21582.5272 / 73 / 149
ShogiBaseline1 working of 1 attempt49.51423.2179 / 92 / 223
XiangqiIntensive1 working of 1 attempt52.81496.5130 / 248 / 120
XiangqiBalanced1 working of 1 attempt47.61456.560 / 337 / 101
XiangqiBaseline1 working of 1 attempt45.91443.1107 / 226 / 165
HexIntensive1 working of 1 attempt62.31596.3321 / 0 / 209
HexBalanced1 working of 1 attempt51.61457.8265 / 0 / 265
HexBaseline1 working of 1 attempt55.31503.4284 / 0 / 246
International DraughtsBalanced1 working of 1 attempt56.01514.5247 / 27 / 228
International DraughtsBaseline1 working of 1 attempt47.81385.6208 / 19 / 275
Game of the AmazonsIntensive1 working of 1 attempt93.72357.1500 / 0 / 30
Game of the AmazonsBalanced1 working of 1 attempt25.21124.2121 / 0 / 409
Game of the AmazonsBaseline1 working of 1 attempt27.91163.8135 / 0 / 395
TumbleweedIntensive1 working of 1 attempt24.91112.933 / 180 / 325
TumbleweedBalanced1 working of 1 attempt39.41272.0134 / 125 / 279
TumbleweedBaseline1 working of 1 attempt92.82260.6496 / 0 / 42

Each row covers all the model's programs in one game and setting. Strength and raw rating are attempt-weighted means; under them, each program's score and the lowest and highest rating describe the programs (not a confidence interval). The combined record adds up their matches: evidence volume, not strength.

Failures and budgets

1 model-code generation failures; 0 provider/infrastructure failures excluded from the generation denominator. Runtime player faults are separate: 0 recorded one-sided faults across contributing program rows. A crash or illegal move pinned on one program also counts as that program's loss in its match record; other faults, such as timeouts, are left out of rating evidence.

Repair/resend outcomes and available costs are included. Exact retry ceilings and execution budgets are not retained in this result set; see the methodology for the public policy. Do not infer a measured setting from a current provider catalog.

Player program results

Max: 88% / 0% / 12% · n 8High: 88% / 12% / 0% · n 8Low: 88% / 12% / 0% · n 8 All three benchmark settings tested

Generation cost

$0.28average estimated cost per game and reasoning setting

Median
$0.25
Range
$0.06–$0.69
Known cost subtotal (incomplete or unverified charges)
$6.65
Cost data
24 combinations · 24 runs
Price source
Provider-reported
Price list
boardgame-list-prices-2026-10-09
Average output
17k tokens
Output range
2.2k–42k tokens
Runs with output-token data
24
How cost is measured

Comparison cost gives equal weight to each model, game, and reasoning-setting combination. Recorded generation attempts include repairs and resends when cost evidence is available; runtime compute and service operation are excluded. Cost and output-token counts use separate telemetry denominators. A list-price estimate is not actual expenditure; mixed evidence is not a uniform standard-price comparison.

Test protocol

  • Combines this model's tested reasoning settings
  • Player-program generation; retained protocol details are under “Test protocol and evidence limits” above