AGC-Bench - analobench: leaderboard
The "analobench" 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. 82 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Sonnet 4.6 | 0.94 | #90 |
| 2 | Claude Opus 4.6 | 0.94 | #63 |
| 3 | Claude Opus 4.7 | 0.94 | #49 |
| 4 | Claude Sonnet 4.5 | 0.85 | #142 |
| 5 | Gemini 2.5 Flash | 0.83 | #245 |
| 6 | Kimi K2.5 | 0.83 | #144 |
| 7 | Claude 3.7 Sonnet | 0.83 | #249 |
| 8 | Claude Opus 4.5 | 0.83 | #83 |
| 9 | DeepSeek V3.2 Exp | 0.83 | #234 |
| 10 | Claude Sonnet 4 | 0.8 | #199 |
| 11 | GPT-5.5 | 0.77 | #31 |
| 12 | pixtral-large-2411 | 0.77 | #478 |
| 13 | Mistral Large 3 | 0.74 | #401 |
| 14 | Gemma 3 12B (IT) | 0.71 | #664 |
| 15 | Llama 4 Maverick | 0.71 | #464 |
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-analobench · How It Works · Data refreshed daily, snapshot 2026-10-20.