Gomoku
15×15 freestyle Gomoku: five or more in a row wins. Rapfi scores every empty cell by win rate; the answer is a cell. See Methodology.
Quality is the mean oracle score of the moves played, × 100. The best move scores 100; an illegal move or no move scores 0. Protocol boardbench-0.2.
Quality vs cost
7 of 7 models
- Anthropic
- OpenAI
- DeepSeek
- Z.ai
- Meta
- Pareto frontier: nothing plotted is both cheaper and better
- Anthropic
- OpenAI
- DeepSeek
- Z.ai
- Meta
Horizontal axis is logarithmic.
Table
| Model | Quality | Legality | Cost | n |
|---|---|---|---|---|
| Muse Spark 1.3High | 46.5 | 100.0 | $0.09 | 3 |
| GLM-5.3 FlashHigh | 34.0 | 100.0 | <$0.01 | 3 |
| Haiku 5.5High | 19.1 | 100.0 | <$0.01 | 3 |
| LunaHigh | 13.5 | 100.0 | <$0.01 | 3 |
| DeepSeek V4.1 FlashHigh | 13.1 | 100.0 | $0.03 | 3 |
| SolHigh | 13.1 | 100.0 | $0.05 | 3 |
| Opus 5.5High | 12.6 | 100.0 | $0.10 | 3 |
Headline is the mean move score × 100. n is answers: puzzles × samples.