CLEM TextMapWorld GraphReasoning — leaderboard
Metric: Game Clemscore (%). Source: huggingface.co. 27 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Kimi K2 (Thinking) | 84.28 |
| 2 | Gemini 3 Flash | 83.08 |
| 3 | Claude Sonnet 4.5 | 82.82 |
| 4 | Claude Sonnet 4.5 (High) | 82.74 |
| 5 | Qwen 3 Max | 82.32 |
| 6 | Kimi K2.5 (Non-reasoning) | 73.16 |
| 7 | GLM-5 (Non-reasoning) | 71.99 |
| 8 | GLM-4.7 | 69.95 |
| 9 | GPT-5.2 | 68.61 |
| 10 | DeepSeek V3.2 | 68.47 |
| 11 | MiniMax-M2.5 | 67.25 |
| 12 | Qwen 2.5 Coder 32B Instruct | 63.24 |
| 13 | DeepSeek V3.2 (Non-reasoning) | 61.18 |
| 14 | Qwen 2.5 72B Instruct | 61.04 |
| 15 | Qwen 3 Next 80B A3B Instruct | 60.58 |
Interactive version: theaggregate.ai/benchmark?slug=clem-textmapworld-graphreasoning · How the rankings work · Data refreshed daily, snapshot 2026-07-22.