SnorkelWordle: leaderboard
Snorkel AI's archived SnorkelWordle benchmark: models play Wordle on 100 low-frequency five-letter English words, receiving partial and exact letter-match feedback after each guess. Win rate is the share of games solved within six guesses; %, higher is better. Frozen roster.
Metric: Win rate (%, 100 games within six guesses). Source: snorkel.ai. Saturation forecast: Estimated already saturated. 21 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5 | 94 | #105 |
| 2 | Grok 4 | 93 | #189 |
| 3 | O3 | 92.9 | #132 |
| 4 | O4 Mini | 91.9 | #185 |
| 5 | GPT-5 Mini | 91 | #193 |
| 6 | Gemini 3 Pro | 91 | #82 |
| 7 | O3 Mini | 90 | #282 |
| 8 | Grok 4 Fast (Reasoning) | 88 | #266 (Grok 4 Fast) |
| 9 | Claude Opus 4 | 85.6 | #170 |
| 10 | Kimi K2 (Thinking) | 85 | #252 (Kimi K2) |
| 11 | Claude Sonnet 4 | 83 | #207 |
| 12 | GPT-OSS-120B | 81.6 | #346 |
| 13 | GPT-5 Nano | 79 | #439 |
| 14 | Gemini 2.5 Pro | 74 | #159 |
| 15 | Grok 3 | 71 | #313 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=snorkelwordle · How It Works · Data refreshed daily, snapshot 2026-11-05.