AGC-Bench - pun_eval: leaderboard
The "pun_eval" 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: Not forecast. 83 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | DeepSeek V3 (0324) | 1.52 | #340 |
| 2 | Mistral Medium 3.1 | 1.42 | #408 |
| 3 | GPT-5.4 Mini | 1.4 | #212 |
| 4 | Claude Opus 4.7 | 1.33 | #49 |
| 5 | DeepSeek V3 Chat | 1.22 | #447 |
| 6 | Ministral 3 8B | 1.22 | #687 |
| 7 | GPT-5.5 | 1.04 | #31 |
| 8 | Gemma 4 26B A4B (IT) | 0.94 | #297 |
| 9 | GPT-5.4 | 0.93 | #79 |
| 10 | Claude Opus 4.5 | 0.9 | #83 |
| 11 | GPT-5.4 Nano | 0.87 | #328 |
| 12 | Qwen 3.5 Flash (02-23) | 0.84 | #185 |
| 13 | NVIDIA Nemotron Nano 9B V2 | 0.83 | #823 |
| 14 | Qwen 3 Next 80B A3B Instruct | 0.8 | #368 |
| 15 | Llama 3 8B Instruct | 0.75 | #1129 |
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-pun-eval · How It Works · Data refreshed daily, snapshot 2026-10-20.