Extended NYT Connections Newest 100
Games · 2026-02-02
Newest-100 split of Lech Mazur's Extended NYT Connections leaderboard. The split is intended to reduce potential training data contamination by scoring the most recent extended puzzles.
Top models (higher is better)
| Model | Score |
|---|---|
| Gemini 3.1 Pro Preview | 97.8 |
| GPT-5.5 | 97.8 |
| GPT-5.6 Sol | 96.8 |
| GPT-5.4 | 96.0 |
| Claude Fable 5 | 95.5 |
| Gemini 3 Pro Preview | 95.2 |
| Opus 4.8 | 95.0 |
| Kimi K3 | 94.5 |
| Gemini 3.5 Flash | 94.2 |
| Opus 4.6 | 91.5 |
| Grok 4.20 | 91.0 |
| Muse Spark 1.1 | 90.0 |
| Kimi K2.6 | 86.5 |
| GPT-5.2 | 85.0 |
| Qwen3.7-Max | 84.8 |
| GLM-5.1 | 84.2 |
| Grok 4.5 | 84.2 |
| Grok 4.1 Fast | 83.8 |
| GLM-5.2 | 80.8 |
| GPT-5.2 Pro | 80.8 |
| GPT-5.6 Terra | 80.8 |
| Sonnet 4.6 | 80.0 |
| Qwen3.6-Max-Preview | 78.5 |
| Sonnet 5 | 78.0 |
| Gemma 4 31B IT | 75.8 |
| GPT-5.6 Luna | 75.8 |
| GLM-5 | 75.5 |
| MiniMax M3 | 74.2 |
| Kimi K2.5 | 73.2 |
| GPT-5.4 Mini | 72.0 |
| Gemini 3 Flash Preview | 71.8 |
| DeepSeek-V4-Pro | 69.2 |
| Qwen3.6 Plus (2026-04-02) | 68.8 |
| Opus 4.5 | 62.5 |
| Grok 4.3 | 60.2 |
| Qwen3.5 122B A10B | 58.8 |
| Qwen3.5 397B A17B | 58.5 |
| Qwen3.5-27B | 58.2 |
| Sonnet 4.5 | 54.0 |
| Haiku 4.5 | 51.5 |