Scores by reasoning setting

Max (Intensive group, tested and included)52.2
High (Balanced group, tested and included)33.9
Low (Baseline group, tested and included)29.4

Every tested setting is included in the overall score.

Available settings and grouping

Available settings

LowHighMax

Checked on 2026-08-21 · Provider documentation

How settings are grouped

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

Scores by game

Go46.6
Chess22.4
Shogi36.6
Xiangqi31.8
Hex48.4
International Draughts47.1
Game of the Amazons45.9
Tumbleweed33.7

Relative scores within each game, across the included settings; they do not compare game difficulty.

What this result shows

Strongest included setting: Max (52.2).

Across included settings, the highest game score is Hex (48.4); the lowest is Chess (22.4). 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: stealth/ox-alpha
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 (102 of 114 indexed attempts)

Generation evidence

Low: 28 contributing player programs across 8 games
High: 26 contributing player programs across 8 games
Max: 25 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; the per-game counts below keep unique code identities separate from admitted generation runs.

Game	Setting	Unique programs	Admitted generation runs	Program score range
Chess	high	3	3	11.5–20.4
Chess	low	3	3	8.7–10.3
Chess	max	2	2	34.1–52.2
Game of the Amazons	high	3	3	18.3–63.8
Game of the Amazons	low	3	3	22.3–51.4
Game of the Amazons	max	3	3	53.0–62.6
Go	high	2	2	49.7–54.6
Go	low	3	3	29.0–44.4
Go	max	3	3	40.4–53.4
Hex	high	3	3	39.8–49.6
Hex	low	3	3	42.9–50.6
Hex	max	3	3	50.0–58.1
International Draughts	high	2	2	9.5–50.1
International Draughts	low	3	3	32.2–55.3
International Draughts	max	3	3	44.3–77.5
Shogi	high	7	7	21.0–61.2
Shogi	low	8	8	10.6–33.9
Shogi	max	5	5	33.4–61.7
Tumbleweed	high	3	3	25.2–47.4
Tumbleweed	low	2	2	16.8–17.7
Tumbleweed	max	3	3	32.9–66.3
Xiangqi	high	3	3	20.9–30.0
Xiangqi	low	3	3	14.8–25.8
Xiangqi	max	3	3	30.2–81.0

Program ranges describe the retained competitors on this release scale, not a statistical estimate of a fresh generation.

Failures and budgets

6 model-code generation failures; 5 provider/infrastructure failures excluded from the generation denominator. Runtime player faults are separate: 183 recorded unilateral faults across contributing program rows, excluded from 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% / 12% / 0% · n 25High: 70% / 17% / 13% · n 30Low: 63% / 30% / 7% · n 30 All three benchmark settings tested

Generation cost

$0.00average estimated cost per game and reasoning setting

Median: $0.00
Range: $0.00–$0.00
Known cost subtotal (incomplete or unverified charges): $0.00
Cost data: 24 combinations · 88 runs
Price source: Provider-reported
Price list: Not recorded for this older data
Average output: 23k tokens
Output range: 1.7k–85k tokens
Runs with output-token data: 88

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

Ox Alpha