AGC-Bench - liveideabench: leaderboard
The "liveideabench" 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 | GPT-5.5 | 1.57 | #31 |
| 2 | Claude Opus 4.6 | 1.51 | #63 |
| 3 | GPT-5.4 | 1.5 | #79 |
| 4 | Claude Opus 4.7 | 1.5 | #49 |
| 5 | GPT-5.1 | 1.5 | #135 |
| 6 | Kimi K2 0905 | 1.48 | #288 |
| 7 | GPT-5.4 Mini | 1.27 | #212 |
| 8 | Mistral Medium 3.1 | 1.24 | #408 |
| 9 | Claude Sonnet 4.5 | 1.15 | #142 |
| 10 | GLM-5.1 | 1.13 | #97 |
| 11 | Gemma 4 26B A4B (IT) | 1.11 | #297 |
| 12 | Claude Opus 4.5 | 1.07 | #83 |
| 13 | GLM-5-Turbo | 1.02 | #119 |
| 14 | Claude Sonnet 4.6 | 0.99 | #90 |
| 15 | nemotron-3-nano-30B-a3B | 0.98 | #662 |
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-liveideabench · How It Works · Data refreshed daily, snapshot 2026-10-20.