Score · relative 0–100: 47.8
Rank: 41
Status: Partial 3/3
Players generated: 2026-10-20
Match evidence updated: 2026-10-24
Results published: 2026-10-26

Compare in charts View player programs Back to leaderboard

Why this result is partial: some games cannot be scored: their only generation attempts failed because of the test harness or the provider and cannot currently be repeated. The rank uses the available reasoning-setting results.

Scores by reasoning setting

High (Intensive group, tested and included)59.6
Medium (Balanced group, tested and included)45.0
None (Baseline group, tested and included)38.9

Every tested setting is included in the overall score.

Available settings and grouping

Available settings

NoneLowMediumHigh

Checked on 2026-10-20 · Provider documentation

How settings are grouped

Baseline: None — tested and included
Balanced: Medium — tested and included
Intensive: High — tested and included

Scores by game

Go42.5
Chess44.4
Shogi1 of 3 failed31.9
Xiangqi40.7
Hex25.3
International Draughts73.3
Game of the Amazons55.7
Tumbleweed50.4

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: High (59.6).

Across included settings, the highest game score is International Draughts (73.3); the lowest is Hex (25.3). 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: inception/mercury-3
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: 131072 (31 of 31 indexed attempts)

Generation evidence

None: 8 contributing player programs across 8 games
Medium: 3 contributing player programs across 3 games
High: 5 contributing player programs across 5 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 1 attempt	50.1	1469.6	118 / 259 / 139
Go	Baseline	1 working of 1 attempt	34.9	1322.7	81 / 179 / 256
Chess	Intensive	1 working of 1 attempt	44.2	1426.8	30 / 348 / 102
Chess	Balanced	1 working of 1 attempt	46.4	1443.9	47 / 335 / 98
Chess	Baseline	1 working of 1 attempt	42.7	1415.3	12 / 370 / 98
Shogi	Intensive	1 working of 1 attempt	62.2	1540.8	231 / 114 / 149
Shogi	Baseline	1 working of 1 attempt	33.7	1284.2	93 / 113 / 288
Xiangqi	Baseline	1 working of 1 attempt	40.7	1402.7	33 / 323 / 142
Hex	Baseline	1 working of 1 attempt	25.3	1100.2	128 / 0 / 402
International Draughts	Intensive	1 working of 1 attempt	79.6	1936.9	386 / 6 / 110
International Draughts	Baseline	1 working of 1 attempt	66.9	1697.3	316 / 5 / 180
Game of the Amazons	Balanced	1 working of 1 attempt	77.9	1851.0	409 / 0 / 121
Game of the Amazons	Baseline	1 working of 1 attempt	33.5	1239.7	164 / 0 / 366
Tumbleweed	Intensive	1 working of 1 attempt	61.7	1547.3	313 / 0 / 225
Tumbleweed	Balanced	1 working of 1 attempt	55.6	1460.6	280 / 0 / 258
Tumbleweed	Baseline	1 working of 1 attempt	33.8	1211.7	168 / 0 / 370

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; 7 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: 40% / 60% / 0% · n 5Medium: 50% / 25% / 25% · n 4None: 88% / 12% / 0% · n 8 All three benchmark settings testedMany programs needed repairMany programs failed

Generation cost

$0.03average estimated cost per game and reasoning setting

Median: $0.04
Range: $0.0016–$0.08
Provider-reported total: $0.69
Cost data: 24 combinations · 24 runs
Price source: Provider-reported
Price list: boardgame-list-prices-2026-10-09
Average output: 80k tokens
Output range: 2.3k–231k 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

Mercury 3