Scores by reasoning setting

High (Intensive group, tested and included)32.9
Medium (Balanced group, tested and included)27.2
Minimal (Baseline group, tested and included)39.2

Every tested setting is included in the overall score.

Available settings and grouping

Available settings

MinimalLowMediumHigh

Checked on 2026-07-24 · Provider documentation

How settings are grouped

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

Scores by game

Go1 of 6 failed31.8
Chess2 of 6 failed28.3
Shogi1 of 6 failed39.1
Xiangqi43.1
Hex15.2
International Draughts37.2
Game of the Amazons1 of 6 failed39.1
Tumbleweed30.8

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: Minimal (39.2).

Across included settings, the highest game score is Xiangqi (43.1); the lowest is Hex (15.2). 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-lite
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 132 indexed attempts)

Generation evidence

Minimal: 15 contributing player programs across 8 games
Medium: 14 contributing player programs across 8 games
High: 14 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.

Game	Reasoning group	Programs	Working-program strength	Raw rating	Combined record (W / D / L)
Go	Intensive	1 working of 2 attempts	35.2	1325.7	122 / 100 / 294
Go	Balanced	2 working of 2 attempts	41.0 program scores 41.0 · 41.0	1383.3 program range 1383–1383	318 / 170 / 544
Go	Baseline	2 working of 2 attempts	36.9 program scores 41.1 · 32.7	1342.2 program range 1300–1384	264 / 195 / 573
Chess	Intensive	1 working of 2 attempts	45.1	1434.1	29 / 359 / 92
Chess	Balanced	1 working of 2 attempts	43.7	1422.7	3 / 397 / 80
Chess	Baseline	2 working of 2 attempts	40.6 program scores 41.7 · 39.5	1399.2 program range 1390–1408	3 / 742 / 214
Shogi	Intensive	2 working of 2 attempts	47.6 program scores 57.7 · 37.4	1407.6 program range 1318–1497	360 / 145 / 483
Shogi	Balanced	2 working of 2 attempts	42.1 program scores 42.7 · 41.6	1359.2 program range 1355–1364	225 / 308 / 455
Shogi	Baseline	1 working of 2 attempts	55.2	1474.2	209 / 88 / 197
Xiangqi	Intensive	2 working of 2 attempts	39.8 program scores 41.4 · 38.1	1395.4 program range 1383–1408	82 / 596 / 318
Xiangqi	Balanced	2 working of 2 attempts	42.5 program scores 43.7 · 41.2	1416.7 program range 1407–1427	76 / 661 / 259
Xiangqi	Baseline	2 working of 2 attempts	47.0 program scores 48.7 · 45.4	1452.1 program range 1439–1465	162 / 579 / 255
Hex	Intensive	2 working of 2 attempts	18.1 program scores 18.9 · 17.2	973.5 program range 957–990	181 / 0 / 879
Hex	Balanced	2 working of 2 attempts	15.8 program scores 16.0 · 15.5	927.9 program range 922–934	157 / 0 / 903
Hex	Baseline	2 working of 2 attempts	11.9 program scores 14.1 · 9.7	839.4 program range 785–893	117 / 0 / 943
International Draughts	Intensive	2 working of 2 attempts	41.4 program scores 62.8 · 20.0	1293.7 program range 960–1627	371 / 22 / 611
International Draughts	Balanced	2 working of 2 attempts	13.9 program scores 14.5 · 13.2	851.6 program range 839–864	114 / 5 / 885
International Draughts	Baseline	2 working of 2 attempts	56.4 program scores 59.7 · 53.2	1521.7 program range 1469–1574	521 / 9 / 474
Game of the Amazons	Intensive	2 working of 2 attempts	44.7 program scores 46.5 · 43.0	1381.9 program range 1360–1404	447 / 0 / 613
Game of the Amazons	Balanced	1 working of 2 attempts	20.8	1056.5	99 / 0 / 431
Game of the Amazons	Baseline	2 working of 2 attempts	62.1 program scores 71.8 · 52.4	1611.7 program range 1478–1746	639 / 0 / 421
Tumbleweed	Intensive	2 working of 2 attempts	31.5	1187.2	38 / 237 / 263
Tumbleweed	Balanced	2 working of 2 attempts	30.3 program scores 31.7 · 28.9	1173.6 program range 1158–1189	119 / 363 / 594
Tumbleweed	Baseline	2 working of 2 attempts	30.7 program scores 32.0 · 29.4	1178.0 program range 1163–1193	125 / 359 / 592

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

5 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: 81% / 6% / 13% · n 16Medium: 88% / 0% / 12% · n 16Minimal: 56% / 38% / 6% · n 16 All three benchmark settings tested

Generation cost

$0.04average estimated cost per game and reasoning setting

Median: $0.04
Range: $0.01–$0.07
Known cost subtotal (incomplete or unverified charges): $1.85
Cost data: 24 combinations · 48 runs
Price source: Provider-reported
Price list: Not recorded for this older data
Average output: 13k tokens
Output range: 1.5k–34k tokens
Runs with output-token data: 48

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

Gemini 3.5 Flash Lite