AGC-Bench - hypobench: leaderboard
The "hypobench" 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 | GPT-5.4 Mini | 0.48 | #212 |
| 2 | NVIDIA Nemotron Nano 9B V2 | 0.44 | #823 |
| 3 | Hermes 4 405B | 0.43 | #334 |
| 4 | Mistral Nemo | 0.43 | #1085 |
| 5 | GPT-5.4 Nano | 0.42 | #328 |
| 6 | Nova 2.0 Lite (Non-reasoning) | 0.42 | #417 (Nova 2.0 Lite) |
| 7 | GPT-5.4 | 0.41 | #79 |
| 8 | DeepSeek V3 (0324) | 0.41 | #340 |
| 9 | Claude Opus 4.5 | 0.4 | #83 |
| 10 | GPT-4.1 | 0.39 | #248 |
| 11 | Nemotron Nano 12B v2 VL | 0.38 | #752 |
| 12 | granite-4.0-h-micro | 0.38 | #1088 |
| 13 | Gemini 2.5 Flash | 0.37 | #245 |
| 14 | GPT-5.1 | 0.37 | #135 |
| 15 | GPT-4o Mini | 0.36 | #599 |
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-hypobench · How It Works · Data refreshed daily, snapshot 2026-10-20.