AGC-Bench - grapheval_review_advisor: leaderboard
The "grapheval_review_advisor" 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 June 2027. 83 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 4.7 | 1.77 | #49 |
| 2 | Claude Opus 4.6 | 1.47 | #63 |
| 3 | Gemini 2.5 Flash | 1.29 | #245 |
| 4 | Claude Opus 4.5 | 1.23 | #83 |
| 5 | Kimi K2.6 | 1.18 | #104 |
| 6 | Kimi K2.5 | 1.16 | #144 |
| 7 | Claude Sonnet 4 | 1.11 | #199 |
| 8 | Claude Sonnet 4.6 | 1.09 | #90 |
| 9 | Claude Sonnet 4.5 | 1.08 | #142 |
| 10 | DeepSeek V3.1 NEX N1 | 1.06 | #327 |
| 11 | GPT-5.4 | 1.03 | #79 |
| 12 | GLM-4.7 | 0.93 | #192 |
| 13 | Qwen 3.5 Flash (02-23) | 0.89 | #185 |
| 14 | Claude Haiku 4.5 | 0.88 | #280 |
| 15 | GLM-4.6 | 0.8 | #252 |
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-grapheval-review-advisor · How It Works · Data refreshed daily, snapshot 2026-10-20.