AGC-Bench - ocw: leaderboard
The "ocw" 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 | ERNIE-4.5-VL-424B-A47B | 1.83 | #391 |
| 2 | Claude Sonnet 4 | 1.79 | #199 |
| 3 | Mistral Large 3 | 1.63 | #401 |
| 4 | Nova Lite (v1) | 1.63 | #830 |
| 5 | DeepSeek V3 (0324) | 1.57 | #340 |
| 6 | Llama 4 Maverick | 1.53 | #464 |
| 7 | nemotron-3-nano-30B-a3B | 1.49 | #662 |
| 8 | Claude Opus 4.5 | 1.48 | #83 |
| 9 | Claude Haiku 4.5 | 1.42 | #280 |
| 10 | Claude Sonnet 4.5 | 1.41 | #142 |
| 11 | minimax-01 | 1.33 | #606 |
| 12 | Gemma 4 26B A4B (IT) | 1.28 | #297 |
| 13 | Claude Opus 4.6 | 1.24 | #63 |
| 14 | Kimi K2.6 | 1.19 | #104 |
| 15 | Claude Sonnet 4.6 | 1.17 | #90 |
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-ocw · How It Works · Data refreshed daily, snapshot 2026-10-20.