# Game Bench Scoring Sheets

These are the public v2 worksheets for the runs currently listed on the Game Bench leaderboard.

Scores are final public consensus values; individual raw Luiz/Athena sub-criterion inputs are not preserved in these generated sheets.

Scores are entered on a 1-5 scale, then published as a percentage:

```
Final % = (raw weighted score / 5.0) * 100
```

| Model | Thinking | Run date | Final score | Sheet | Game |
|-------|----------|----------|---------------|-------|------|
| Claude Fable 5.0 | high | 2026-06-09 | 95% | [sheet](./claude-fable50.md) | [play](../games/claude-fable50/) |
| Claude Opus 4.8 | xhigh | 2026-06-02 | 89.6% | [sheet](./claude-opus48.md) | [play](../games/claude-opus48/) |
| GPT-5 Codex | xhigh | 2026-06-02 | 83.9% | [sheet](./gpt5-codex.md) | [play](../games/gpt5-codex/) |
| GLM 5.2 | high | 2026-06-22 | 81.5% | [sheet](./glm-52.md) | [play](../games/glm-52/) |
| GLM 5.1 | high | 2026-06-22 | 77% | [sheet](./glm-51.md) | [play](../games/glm-51/) |
| Gemini 3.5 Flash | high | 2026-06-02 | 74.5% | [sheet](./gemini-35-flash.md) | [play](../games/gemini-35-flash/) |
| Kimi K2.6 | high | 2026-06-02 | 66.9% | [sheet](./kimi-k26.md) | [play](../games/kimi-k26/) |
| MiniMax M3 | high | 2026-06-01 | 63.8% | [sheet](./minimax-m3-v3-high.md) | [play](../games/minimax-m3-v3-high/) |
| MiMo V2.5 Pro | medium | 2026-05-29 | 62.1% | [sheet](./v3-medium.md) | [play](../games/v3-medium/) |
| Qwen 3.7 Plus | high | 2026-06-14 | 60.5% | [sheet](./qwen-37-plus.md) | [play](../games/qwen-37-plus/) |
| MiMo V2.5 Pro | max | 2026-06-01 | 56.9% | [sheet](./v3-max.md) | [play](../games/v3-max/) |
| Kimi K2.7 Code | high | 2026-06-12 | 51.4% | [sheet](./kimi-k27-code.md) | [play](../games/kimi-k27-code/) |
| MiMo V2.5 Pro | high | 2026-05-29 | 45.3% | [sheet](./v3-high.md) | [play](../games/v3-high/) |
| DeepSeek V4 Pro | high | 2026-06-22 | 38.9% | [sheet](./ds-v4-pro.md) | [play](../games/ds-v4-pro/) |