Scores by reasoning setting

High (Intensive group, tested and included)33.4
Budgetednot included (Intensive group, tested, not included)33.4
Medium (Balanced group, tested and included)36.2
Budgetednot included (Balanced group, tested, not included)36.2
None (Baseline group, tested and included)33.7
Historicalnot included (Not grouped group, tested, not included)32.0

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

Available settings and grouping

Available settings

Thinking field omittedLow effort + 8,000 thinking tokensMedium effort + 16,000 thinking tokensHigh effort + 32,000 thinking tokens

Checked on 2026-07-21 · Provider documentation

How settings are grouped

Baseline: Thinking field omitted — tested and included
Balanced: Medium effort + 16,000 thinking tokens — tested and included
Intensive: High effort + 32,000 thinking tokens — tested and included

Scores by game

Go47.0
Chess9.6
Shogi25.2
Xiangqi13.8
Hex45.5
International Draughts49.4
Game of the Amazons41.2
Tumbleweed43.8

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

What this result shows

Strongest included setting: Medium effort + 16,000 thinking tokens (36.2).

Across included settings, the highest game score is International Draughts (49.4); the lowest is Chess (9.6). 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: claude-opus-4-5
Provider route (catalog): anthropic
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: 64000 (16 of 24 indexed attempts)

Generation evidence

Thinking field omitted: 8 contributing player programs across 8 games
Medium effort + 16,000 thinking tokens: 8 contributing player programs across 8 games
High effort + 32,000 thinking tokens: 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; 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	1	1	9.2–9.2
Chess	legacy_shared_thinking_budget	2	2	8.8–10.3
Chess	medium	1	1	9.0–9.0
Chess	none	1	1	10.8–10.8
Game of the Amazons	high	1	1	44.8–44.8
Game of the Amazons	legacy_shared_thinking_budget	2	2	49.6–54.3
Game of the Amazons	medium	1	1	41.8–41.8
Game of the Amazons	none	1	1	37.0–37.0
Go	high	1	1	56.2–56.2
Go	legacy_shared_thinking_budget	2	2	26.1–32.0
Go	medium	1	1	48.8–48.8
Go	none	1	1	36.0–36.0
Hex	high	1	1	47.1–47.1
Hex	legacy_shared_thinking_budget	2	2	41.3–43.3
Hex	medium	1	1	44.2–44.2
Hex	none	1	1	45.3–45.3
International Draughts	high	1	1	53.6–53.6
International Draughts	legacy_shared_thinking_budget	2	2	46.2–47.7
International Draughts	medium	1	1	49.7–49.7
International Draughts	none	1	1	44.9–44.9
Shogi	high	1	1	26.5–26.5
Shogi	legacy_shared_thinking_budget	1	1	28.0–28.0
Shogi	medium	1	1	27.0–27.0
Shogi	none	1	1	22.0–22.0
Tumbleweed	high	1	1	20.0–20.0
Tumbleweed	legacy_shared_thinking_budget	2	2	32.9–53.0
Tumbleweed	medium	1	1	57.0–57.0
Tumbleweed	none	1	1	54.2–54.2
Xiangqi	high	1	1	9.5–9.5
Xiangqi	legacy_shared_thinking_budget	2	2	9.0–10.5
Xiangqi	medium	1	1	12.2–12.2
Xiangqi	none	1	1	19.6–19.6

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

Failures and budgets

0 model-code generation failures; 0 provider/infrastructure failures excluded from the generation denominator. Runtime player faults are separate: 1 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

High effort + 32,000 thinking tokens: 100% / 0% / 0% · n 8Medium effort + 16,000 thinking tokens: 100% / 0% / 0% · n 8Thinking field omitted: 100% / 0% / 0% · n 8 All three benchmark settings tested

Generation cost

$0.25average estimated cost per game and reasoning setting

Median: $0.21
Range: $0.10–$0.51
Known cost subtotal (incomplete or unverified charges): $11.20
Cost data: 24 combinations · 40 runs
Price source: Standard list-price estimate
Price list: boardgame-list-prices-2026-07-12 + boardgame-list-prices-2026-08-08
Average output: 10.0k tokens
Output range: 2.1k–28k 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

Claude Opus 4.5