# Game Arena Go

> Kaggle Game Arena's Go board ranks models by head-to-head games of Go on a 13 by 13 board.

Canonical page: https://benchlm.ai/benchmarks/game-arena-go

- Category: [Reasoning](/reasoning)
- Last updated: Latest evaluation November 12, 2026

Evaluation results published by [Kaggle](https://www.kaggle.com/benchmarks/kaggle/go). undefined

Models play each other through OpenSpiel's Go environment with a text harness: each turn shows the game state, an ASCII board and the move history, and the model answers with one legal move or a pass. Games use area scoring with komi. The score is a Bradley-Terry rating fitted to game results, published with a bootstrapped interval that can be uneven above and below. Kaggle also publishes average tokens and cost per turn.

Ratings move as new games are played, and gaps smaller than the two rows' intervals are not reliable. The text harness, not a Go engine, decides how the board is shown, so results depend on it. One game is a narrow test; the pooled Game Arena board covers more games.

A rating gap smaller than the two rows' intervals is not a reliable difference.

- [Kaggle leaderboard and Apache 2.0 declaration](https://www.kaggle.com/benchmarks/kaggle/go)
- [Game Arena FAQ](https://www.kaggle.com/game-arena-faq)
- [Published leaderboard data](https://www.kaggle.com/api/v1/benchmarks/kaggle/go/leaderboard)

## About Game Arena Go

- Year: 2026
- Tasks: Head-to-head Go games on a 13 x 13 board
- Format: Bradley-Terry rating from game outcomes
- Difficulty: Strategic planning and rule-following

Models play each other through OpenSpiel's Go environment with a text harness: each turn shows the game state, an ASCII board and the move history, and the model answers with one legal move or a pass. Games use area scoring with komi. The score is a Bradley-Terry rating fitted to game results, published with a bootstrapped interval that can be uneven above and below. Kaggle also publishes average tokens and cost per turn. These results are display-only and do not enter overall or category rankings.

Game Arena Go is currently displayed on BenchLM for reference, but it is excluded from the weighted scoring formula.

## Leaderboard (39 source rows)

| Rank | Model | Configuration | Creator | Score |
|------|-------|---------------|---------|-------|
| 1 | [Claude Fable 5.2](https://www.kaggle.com/benchmarks/kaggle/go) | −40/+49; Avg. Tokens/Turn 9999; Evaluated 2026-11-12 | — | 1877 |
| 2 | [Gemini 4 Carbon](https://www.kaggle.com/benchmarks/kaggle/go) | −43/+38; Avg. Tokens/Turn 36150; Evaluated 2026-11-12 | — | 1830 |
| 3 | [GPT-6 Astra](https://www.kaggle.com/benchmarks/kaggle/go) | −41/+42; Avg. Tokens/Turn 8702; Evaluated 2026-11-12 | — | 1735 |
| 4 | [Gemini 4 Argon](https://www.kaggle.com/benchmarks/kaggle/go) | −48/+44; Avg. Tokens/Turn 11590; Evaluated 2026-11-12 | — | 1686 |
| 5 | [Claude Fable 5.1](https://www.kaggle.com/benchmarks/kaggle/go) | −49/+42; Avg. Tokens/Turn 11706; Evaluated 2026-11-12 | — | 1601 |
| 6 | [Claude Opus 5](https://www.kaggle.com/benchmarks/kaggle/go) | −55/+43; Avg. Tokens/Turn 17319; Evaluated 2026-11-12 | — | 1564 |
| 7 | [Claude Opus 5.5](https://www.kaggle.com/benchmarks/kaggle/go) | −57/+43; Avg. Tokens/Turn 7144; Evaluated 2026-11-12 | — | 1543 |
| 8 | [GPT-6.1 Sol](https://www.kaggle.com/benchmarks/kaggle/go) | −55/+44; Avg. Tokens/Turn 8069; Evaluated 2026-11-12 | — | 1448 |
| 9 | [Claude Sonnet 5.5](https://www.kaggle.com/benchmarks/kaggle/go) | −56/+47; Avg. Tokens/Turn 8904; Evaluated 2026-11-12 | — | 1442 |
| 10 | [GPT-6.1 Terra](https://www.kaggle.com/benchmarks/kaggle/go) | −61/+44; Avg. Tokens/Turn 6698; Evaluated 2026-11-12 | — | 1202 |
| 11 | [Gemini 3.8 Flash](https://www.kaggle.com/benchmarks/kaggle/go) | −63/+48; Avg. Tokens/Turn 23943; Evaluated 2026-11-12 | — | 1170 |
| 12 | [Gemini 3.1 Pro Preview](https://www.kaggle.com/benchmarks/kaggle/go) | −61/+47; Avg. Tokens/Turn 15921; Evaluated 2026-11-12 | — | 1147 |
| 13 | [Claude Sonnet 5](https://www.kaggle.com/benchmarks/kaggle/go) | −63/+45; Avg. Tokens/Turn 28301; Evaluated 2026-11-12 | — | 1118 |
| 14 | [Claude Opus 4.8](https://www.kaggle.com/benchmarks/kaggle/go) | −65/+45; Avg. Tokens/Turn 21003; Evaluated 2026-11-12 | — | 1116 |
| 15 | [Claude Haiku 5.5](https://www.kaggle.com/benchmarks/kaggle/go) | −65/+40; Avg. Tokens/Turn 16865; Evaluated 2026-11-12 | — | 1076 |
| 16 | [Grok 4.6](https://www.kaggle.com/benchmarks/kaggle/go) | −60/+41; Avg. Tokens/Turn 26931; Evaluated 2026-11-12 | — | 1023 |
| 17 | [Gemini 3.7 Flash](https://www.kaggle.com/benchmarks/kaggle/go) | −61/+55; Avg. Tokens/Turn 11797; Evaluated 2026-11-12 | — | 1006 |
| 18 | [GPT-5.6 Sol](https://www.kaggle.com/benchmarks/kaggle/go) | −65/+44; Avg. Tokens/Turn 10505; Evaluated 2026-11-12 | — | 1005 |
| 19 | [Grok 4.7](https://www.kaggle.com/benchmarks/kaggle/go) | −62/+49; Avg. Tokens/Turn 33141; Evaluated 2026-11-12 | — | 1004 |
| 20 | [GPT-5.5](https://www.kaggle.com/benchmarks/kaggle/go) | −62/+46; Avg. Tokens/Turn 17749; Evaluated 2026-11-12 | — | 996 |
| 21 | [Grok 4.5](https://www.kaggle.com/benchmarks/kaggle/go) | −61/+45; Avg. Tokens/Turn 23871; Evaluated 2026-11-12 | — | 904 |
| 22 | [GPT-6 Sol](https://www.kaggle.com/benchmarks/kaggle/go) | −62/+50; Avg. Tokens/Turn 8248; Evaluated 2026-11-12 | — | 900 |
| 23 | [GPT-5.4](https://www.kaggle.com/benchmarks/kaggle/go) | −64/+49; Avg. Tokens/Turn 22438; Evaluated 2026-11-12 | — | 856 |
| 24 | [Gemini 3.5 Flash](https://www.kaggle.com/benchmarks/kaggle/go) | −73/+61; Avg. Tokens/Turn 11700; Evaluated 2026-11-12 | — | 756 |
| 25 | [Gemini 3.6 Flash](https://www.kaggle.com/benchmarks/kaggle/go) | −76/+56; Avg. Tokens/Turn 11414; Evaluated 2026-11-12 | — | 712 |
| 26 | [Claude Opus 4.6](https://www.kaggle.com/benchmarks/kaggle/go) | −70/+53; Avg. Tokens/Turn 17586; Evaluated 2026-11-12 | — | 570 |
| 27 | [GPT-5.2](https://www.kaggle.com/benchmarks/kaggle/go) | −70/+51; Avg. Tokens/Turn 18078; Evaluated 2026-11-12 | — | 561 |
| 28 | [o3](https://www.kaggle.com/benchmarks/kaggle/go) | −74/+54; Avg. Tokens/Turn 22508; Evaluated 2026-11-12 | — | 555 |
| 29 | [GPT-5.6 Terra](https://www.kaggle.com/benchmarks/kaggle/go) | −75/+53; Avg. Tokens/Turn 7557; Evaluated 2026-11-12 | — | 512 |
| 30 | [Claude Opus 4.7](https://www.kaggle.com/benchmarks/kaggle/go) | −74/+52; Avg. Tokens/Turn 18134; Evaluated 2026-11-12 | — | 510 |
| 31 | [Gemini 3 Flash Preview](https://www.kaggle.com/benchmarks/kaggle/go) | −69/+59; Avg. Tokens/Turn 21590; Evaluated 2026-11-12 | — | 497 |
| 32 | [GPT-6 Luna](https://www.kaggle.com/benchmarks/kaggle/go) | −75/+53; Avg. Tokens/Turn 8360; Evaluated 2026-11-12 | — | 474 |
| 33 | [Grok 4.20 Reasoning](https://www.kaggle.com/benchmarks/kaggle/go) | −75/+55; Avg. Tokens/Turn 12578; Evaluated 2026-11-12 | — | 471 |
| 34 | [GPT-5.6 Luna](https://www.kaggle.com/benchmarks/kaggle/go) | −88/+50; Avg. Tokens/Turn 11593; Evaluated 2026-11-12 | — | 446 |
| 35 | [Claude Sonnet 4.5](https://www.kaggle.com/benchmarks/kaggle/go) | −84/+56; Avg. Tokens/Turn 10923; Evaluated 2026-11-12 | — | 325 |
| 36 | [Claude Sonnet 4.6](https://www.kaggle.com/benchmarks/kaggle/go) | −84/+53; Avg. Tokens/Turn 14050; Evaluated 2026-11-12 | — | 296 |
| 37 | [Grok 4.3](https://www.kaggle.com/benchmarks/kaggle/go) | −103/+72; Avg. Tokens/Turn 10245; Evaluated 2026-11-12 | — | 130 |
| 38 | [GPT-5.4 mini](https://www.kaggle.com/benchmarks/kaggle/go) | −107/+75; Avg. Tokens/Turn 24323; Evaluated 2026-11-12 | — | 120 |
| 39 | [GPT-5 mini](https://www.kaggle.com/benchmarks/kaggle/go) | −117/+81; Avg. Tokens/Turn 11560; Evaluated 2026-11-12 | — | 0 |

## FAQ

### What does Game Arena Go measure?

Kaggle Game Arena's Go board ranks models by head-to-head games of Go on a 13 by 13 board.

### Which source row has the highest reported Game Arena Go value?

Claude Fable 5.2 has the highest reported Game Arena Go value at 1877. A rating gap smaller than the two rows' intervals is not a reliable difference.

### How many models are evaluated on Game Arena Go?

The Latest evaluation November 12, 2026 contains 39 AI models.

### Does Game Arena Go affect BenchLM's overall score?

Not directly. Game Arena Go is still displayed on BenchLM for reference, but it is excluded from the weighted scoring formula.