AGC-Bench - fig_qa: leaderboard
The "fig_qa" 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 September 2027. 83 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 4.7 | 2.5 | #49 |
| 2 | jamba-large-1.7 | 2.44 | #525 |
| 3 | Gemma 4 31B (IT) | 2.38 | #176 |
| 4 | GLM-5.1 | 2.38 | #97 |
| 5 | GLM-5 | 2.38 | #141 |
| 6 | GLM-4.7 | 2.38 | #192 |
| 7 | GLM-4.6 | 2.38 | #252 |
| 8 | GLM-5-Turbo | 2.38 | #119 |
| 9 | Claude Sonnet 4.6 | 1.95 | #90 |
| 10 | Claude Opus 4.6 | 1.95 | #63 |
| 11 | ERNIE 4.5 300B A47B | 1.89 | #458 |
| 12 | Gemma 4 26B A4B (IT) | 1.21 | #297 |
| 13 | Claude Opus 4.5 | 0.84 | #83 |
| 14 | DeepSeek V3.1 NEX N1 | 0.78 | #327 |
| 15 | GPT-5.4 Mini | 0.47 | #212 |
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-fig-qa · How It Works · Data refreshed daily, snapshot 2026-10-20.