AGC-Bench - cpers: leaderboard
The "cpers" 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 March 2027. 83 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Mistral Medium 3.1 | 1.65 | #408 |
| 2 | Kimi K2 0905 | 1.4 | #288 |
| 3 | Gemma 4 26B A4B (IT) | 1.23 | #297 |
| 4 | Claude 3.7 Sonnet | 1.17 | #249 |
| 5 | GLM-5-Turbo | 1.14 | #119 |
| 6 | Mistral Large 3 | 1.14 | #401 |
| 7 | GLM-4.7 | 1.13 | #192 |
| 8 | Claude Opus 4.6 | 1.09 | #63 |
| 9 | DeepSeek V3.1 NEX N1 | 1.05 | #327 |
| 10 | Kimi K2.6 | 1 | #104 |
| 11 | Qwen 3.5 Flash (02-23) | 0.95 | #185 |
| 12 | GLM-5.1 | 0.94 | #97 |
| 13 | DeepSeek V3.2 Exp | 0.93 | #234 |
| 14 | DeepSeek V3.1 Terminus | 0.92 | #220 |
| 15 | Claude Opus 4.7 | 0.92 | #49 |
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-cpers · How It Works · Data refreshed daily, snapshot 2026-10-20.