NYT Connections Older Models: leaderboard
Historical or incomplete Extended NYT Connections runs with up to four extra decoy words per puzzle. Scores average quadratic partial credit for exact groups; each model was evaluated on its own puzzle sample, smaller than the full 940-puzzle main board.
Metric: Score (%). Source: github.com. Saturation forecast: Around January 2027. 114 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3.8 Max (0902) | 88.8 | #29 |
| 2 | Grok 4 Fast (Reasoning) | 88 | #254 (Grok 4 Fast) |
| 3 | Grok 4 | 87.8 | #181 |
| 4 | GPT-5 Pro | 77 | #81 |
| 5 | O3 (High) | 70.5 | #124 (O3) |
| 6 | GPT-5 (High) | 67.8 | #98 (GPT-5) |
| 7 | O4 Mini (High) | 64 | #180 (O4 Mini) |
| 8 | O3 (Medium) | 63 | #124 (O3) |
| 9 | GPT-5 (Medium) | 61.8 | #98 (GPT-5) |
| 10 | O1 | 60.1 | #183 |
| 11 | GPT-5.1 (High) | 59.7 | #135 (GPT-5.1) |
| 12 | O4 Mini (Medium) | 58.4 | #180 (O4 Mini) |
| 13 | GPT-5 Mini (Medium) | 54.9 | #186 (GPT-5 Mini) |
| 14 | GPT-5 (Low) | 53.2 | #98 (GPT-5) |
| 15 | GPT-5.1 (Medium) | 51.2 | #135 (GPT-5.1) |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=nyt-connections-older-models · How It Works · Data refreshed daily, snapshot 2026-10-20.