DuelLabGameBench 2
Menu

Model results · Google

Gemini 3.5 Flash

Rank 34 of 66 in Overall Models. All three benchmark settings tested.

Score · relative 0–100
51.1
Rank
34
Status
Complete 3/3
Rank change
0
Players generated
2026-07-21 – 2026-09-03
Match evidence updated
2026-10-26
Results published
Compare in charts View player programs Back to leaderboard

Scores by reasoning setting

  • High (Intensive group, tested and included)64.1
  • Medium (Balanced group, tested and included)57.4
  • Minimal (Baseline group, tested and included)31.9
  • Lownot included (Baseline group, tested, not included)31.9
  • Minimalnot included (Baseline group, tested, not included)35.5
  • Lownot included (Baseline group, tested, not included)28.2

Settings marked “not included” were tested but are left out of the overall score.

Available settings and grouping

Available settings

MinimalLowMediumHigh

Checked on 2026-07-21 · Provider documentation

How settings are grouped

  • Baseline: Minimal — tested and included
  • Balanced: Medium — tested and included
  • Intensive: High — tested and included

Scores by game

  • Go71.1
  • Chess1 of 4 failed42.4
  • Shogi1 of 4 failed50.9
  • Xiangqi45.4
  • Hex45.1
  • International Draughts63.1
  • Game of the Amazons50.5
  • Tumbleweed40.3

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: High (64.1).

Across included settings, the highest game score is Go (71.1); the lowest is Tumbleweed (40.3). 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
google/gemini-3.5-flash
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
Not retained (0 of 32 indexed attempts)

Generation evidence

  • Minimal: 8 contributing player programs across 8 games
  • Medium: 8 contributing player programs across 8 games
  • High: 8 contributing player programs across 8 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 attempt90.51996.6450 / 21 / 45
GoBalanced1 working of 1 attempt73.41711.1361 / 14 / 141
GoBaseline2 working of 2 attempts49.5 program scores 60.2 · 38.71464.5 program range 1361–1568397 / 185 / 450
ChessIntensive1 working of 1 attempt62.81575.4161 / 265 / 54
ChessBalanced1 working of 1 attempt44.61430.13 / 406 / 71
ChessBaseline1 working of 2 attempts39.61391.32 / 361 / 117
ShogiIntensive1 working of 1 attempt71.01637.6312 / 43 / 139
ShogiBalanced1 working of 1 attempt70.21627.9304 / 49 / 140
ShogiBaseline1 working of 2 attempts23.11179.166 / 69 / 359
XiangqiIntensive1 working of 1 attempt40.01397.159 / 264 / 175
XiangqiBalanced1 working of 1 attempt54.31508.4120 / 283 / 95
XiangqiBaseline2 working of 2 attempts41.9 program scores 43.9 · 39.91412.2 program range 1396–142883 / 636 / 277
HexIntensive1 working of 1 attempt52.01462.5267 / 0 / 263
HexBalanced1 working of 1 attempt67.21667.8347 / 0 / 183
HexBaseline2 working of 2 attempts16.1 program scores 17.8 · 14.5934.9 program range 902–968161 / 0 / 899
International DraughtsIntensive1 working of 1 attempt75.51856.9363 / 7 / 132
International DraughtsBalanced1 working of 1 attempt74.51836.2357 / 7 / 138
International DraughtsBaseline2 working of 2 attempts39.4 program scores 60.2 · 18.61260.5 program range 938–1583359 / 6 / 639
Game of the AmazonsIntensive1 working of 1 attempt52.01472.9263 / 0 / 267
Game of the AmazonsBalanced1 working of 1 attempt54.21500.6275 / 0 / 255
Game of the AmazonsBaseline2 working of 2 attempts45.3 program scores 54.6 · 36.01388.6 program range 1272–1505454 / 0 / 606
TumbleweedIntensive1 working of 1 attempt68.81664.7315 / 75 / 148
TumbleweedBalanced1 working of 1 attempt20.61060.867 / 69 / 402
TumbleweedBaseline2 working of 2 attempts31.4 program scores 32.3 · 30.51186.1 program range 1176–1196123 / 378 / 575

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

2 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

High: 63% / 37% / 0% · n 8Medium: 75% / 25% / 0% · n 8Minimal: 69% / 19% / 12% · n 16 All three benchmark settings tested

Generation cost

$0.18average estimated cost per game and reasoning setting

Median
$0.19
Range
$0.04–$0.51
Known cost subtotal (incomplete or unverified charges)
$4.98
Cost data
24 combinations · 32 runs
Price source
Combined: Provider-reported + Standard list-price estimate
Price list
boardgame-list-prices-2026-08-08
Average output
19k tokens
Output range
1.8k–61k 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