By model and setting

Each model and reasoning setting appears once. If a model produced more than one player program for this game, the row stands for all of them: Programs counts the working attempts out of the counted attempts, Score and Raw rating average the programs (weighted by attempts; the small range is the lowest and highest program rating), and the combined record adds up their matches. Score is relative within this game (0–100). Scores here are working-program strength: failed generation attempts are left out and count as 0 only in a model's headline score.

GB2-03: Shogi - Reasoning Variants
Rank Model Reasoning Score Programs Raw rating Combined record (W / D / L)
1Claude Fable 5.2XHigh98.41 working of 1 attempt2502.3559 / 4 / 7
2GPT-6 AstraXHigh98.31 working of 1 attempt2493.5557 / 11 / 4
3Gemini 4 CarbonHigh98.01 working of 1 attempt2454.0401 / 3 / 8
4GPT-6.1 SolXHigh97.71 working of 1 attempt2422.4558 / 1 / 13
5Claude Opus 5.5XHigh97.61 working of 1 attempt2415.6556 / 2 / 14
6Gemini 4 HydrogenXHigh97.21 working of 1 attempt2376.6406 / 5 / 9
7Pareto 26.10Default96.41 working of 1 attempt2311.6546 / 3 / 21
8Gemini 4 CarbonMedium96.21 working of 1 attempt2296.9547 / 7 / 22
9GPT-6.1 SolMedium96.01 working of 1 attempt2281.5543 / 4 / 25
10Claude Sonnet 5.5XHigh94.81 working of 1 attempt2209.3538 / 0 / 34
11GPT-6.1 Terra ProXHigh94.61 working of 1 attempt2199.9529 / 10 / 31
12Claude Fable 5.1XHigh94.61 working of 1 attempt2198.5536 / 2 / 34
13Claude Fable 5.2Low94.31 working of 1 attempt2181.5532 / 2 / 36
14GPT-6 AstraMedium93.91 working of 1 attempt2163.7528 / 8 / 36
15Claude Opus 5XHigh93.32 working of 2 attempts2135.3 program range 2105–21661048 / 13 / 85
16GPT-6 SolXHigh93.21 working of 1 attempt2130.2523 / 6 / 43
17Claude Fable 5.2Medium93.01 working of 1 attempt2118.9504 / 43 / 23
18Gemini 4 ArgonHigh92.11 working of 1 attempt2081.7493 / 50 / 27
19Pareto 26.10 PreviewDefault92.11 working of 1 attempt2081.5515 / 7 / 50
20Claude Opus 5.5Medium92.11 working of 1 attempt2081.0506 / 28 / 38
21Claude Fable 5.1Low91.61 working of 1 attempt2061.0515 / 3 / 56
22Claude Fable 5.1Medium91.01 working of 1 attempt2040.7505 / 15 / 52
23GPT-6.1 TerraXHigh89.01 working of 1 attempt1972.1462 / 65 / 43
24Nex N2.6Low87.61 working of 1 attempt1933.9351 / 7 / 54
25GPT-6.1 Terra ProNone87.61 working of 1 attempt1932.6483 / 8 / 79

Individual player programs

Each row is one accepted player program for GB2-03: Shogi; a model can have more than one. Score is the 0–100 result in this game, and its rating and W / D / L record show how it played. Scores are working-program strength: failed generation attempts have no row and count only in a model's headline score.

290 player programs.

ModelReasoningStatusScoreRatingW / D / LMatchesOpponentsNotesGenerated
Claude Fable 5.2E-53600b7725dfeb1bXHighRated Interval validated98.42502559 / 4 / 7570 / 572285—2026-11-09First try
GPT-6 AstraE-18e6aa627f4b1222XHighRated Interval validated98.32494557 / 11 / 4572 / 574286—2026-09-05First try
Gemini 4 CarbonE-06596dc7382fab4fHighRated Interval validated98.02454401 / 3 / 8412 / 414206—2026-11-12First try
GPT-6.1 SolE-27bd4aa164eef135XHighRated Interval validated97.72422558 / 1 / 13572 / 573286—2026-10-01First try
Claude Opus 5.5E-0ac738bb34b1e7c8XHighRated Interval validated97.62416556 / 2 / 14572 / 574286—2026-09-22First try
Gemini 4 HydrogenE-a064debfcb33dc65XHighRated Interval validated97.22377406 / 5 / 9420 / 422210—2026-11-12Repaired
Pareto 26.10E-3f0521308d4debcbDefaultRated Interval validated96.42312546 / 3 / 21570 / 572285—2026-10-21First try
Gemini 4 CarbonE-1e3ddf2284bebdfdMediumRated Interval validated96.22297547 / 7 / 22576 / 578288—2026-11-12First try
GPT-6.1 SolE-fb177ebd31849b47MediumRated Interval validated96.02281543 / 4 / 25572 / 574286—2026-10-01First try
Claude Sonnet 5.5E-4a7793fa8a4833beXHighRated Interval validated94.82209538 / 0 / 34572 / 574286—2026-09-30First try
GPT-6.1 Terra ProE-f5fa770829216201XHighRated Interval validated94.62200529 / 10 / 31570 / 572285—2026-11-01First try
Claude Fable 5.1E-b53393b05159fc95XHighRated Interval validated94.62198536 / 2 / 34572 / 574286—2026-09-08First try
Claude Fable 5.2E-c1d0091908115770LowRated Interval validated94.32182532 / 2 / 36570 / 572285—2026-11-09First try
Claude Fable 5.1E-9f7c077c66189b69XHighRated Interval validated94.02169531 / 4 / 37572 / 574286—2026-09-03First try
Claude Opus 5E-7caa6c5081d57676XHighRated Interval validated94.02166529 / 8 / 37574 / 576287—2026-07-25First try
GPT-6 AstraE-4633465f951f8caeMediumRated Interval validated93.92164528 / 8 / 36572 / 573286—2026-09-05First try
GPT-6 SolE-049711eb0d28bfa8XHighRated Interval validated93.22130523 / 6 / 43572 / 574286—2026-09-22First try
Claude Fable 5.2E-be0fb5f0eb971c6eMediumRated Interval validated93.02119504 / 43 / 23570 / 571285—2026-11-09First try
Claude Opus 5E-38759212c025e8ebXHighRated Interval validated92.72105519 / 5 / 48572 / 574286—2026-07-25First try
Gemini 4 ArgonE-674c6806d0094ec2HighRated Interval validated92.12082493 / 50 / 27570 / 572285—2026-10-21Repaired
Pareto 26.10 PreviewE-74f4f6139b879936DefaultRated Interval validated92.12082515 / 7 / 50572 / 574286—2026-10-09First try
Claude Opus 5.5E-8b93ed37e8717e57MediumRated Interval validated92.12081506 / 28 / 38572 / 574286—2026-09-22First try
Claude Fable 5.1E-529bc866427d95d5LowRated Interval validated91.62061515 / 3 / 56574 / 576287—2026-09-03First try
Claude Fable 5.1E-5c6dabbae6a319cfMediumRated Interval validated91.02041505 / 15 / 52572 / 574286—2026-09-03First try
GPT-6.1 TerraE-f53bdd0e9e794a81XHighRated Interval validated89.01972462 / 65 / 43570 / 571285—2026-11-01First try