AGC-Bench - riddlesense: leaderboard
The "riddlesense" subtask of AGC-Bench. A task from AGC-Bench, evaluating models on open-ended creative and figurative-language generation — humor, metaphor, analogy, storytelling and brainstorming — typically graded by LLM-as-judge.
Metric: Dataset z-score. Source: huggingface.co. Saturation forecast: Around August 2027. 81 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude 3.7 Sonnet | 1.25 | #249 |
| 2 | GLM-5-Turbo | 1.15 | #119 |
| 3 | GLM-5 | 1.15 | #141 |
| 4 | Kimi K2.5 | 1.08 | #144 |
| 5 | Gemma 4 31B (IT) | 1.06 | #176 |
| 6 | Gemma 4 26B A4B (IT) | 1.06 | #297 |
| 7 | GPT-5.4 | 1.01 | #79 |
| 8 | GPT-5.5 | 1 | #31 |
| 9 | GLM-4.7 | 0.99 | #192 |
| 10 | Hermes 4 405B | 0.98 | #334 |
| 11 | ERNIE-4.5-VL-424B-A47B | 0.95 | #391 |
| 12 | Qwen 3 Next 80B A3B Instruct | 0.9 | #368 |
| 13 | Mistral Large 3 | 0.88 | #401 |
| 14 | GPT-4o | 0.82 | #343 |
| 15 | GPT-5.1 | 0.8 | #135 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=agc-bench-riddlesense · How It Works · Data refreshed daily, snapshot 2026-10-20.