AGC-Bench - Problem Solving: leaderboard
The "Problem Solving" 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: JRT z-score. Source: huggingface.co. Saturation forecast: Around April 2027. 83 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Kimi K2.5 | 1.02 | #144 |
| 2 | Claude Opus 4.7 | 0.92 | #49 |
| 3 | GPT-5.5 | 0.78 | #31 |
| 4 | Hermes 4 405B | 0.73 | #334 |
| 5 | GLM-4.7 | 0.72 | #192 |
| 6 | Claude Sonnet 4.5 | 0.7 | #142 |
| 7 | Claude Opus 4.5 | 0.69 | #83 |
| 8 | GLM-5 | 0.68 | #141 |
| 9 | GPT-4o | 0.62 | #343 |
| 10 | Claude Sonnet 4.6 | 0.61 | #90 |
| 11 | GLM-5-Turbo | 0.61 | #119 |
| 12 | Kimi K2.6 | 0.56 | #104 |
| 13 | Kimi K2 0905 | 0.53 | #288 |
| 14 | Mistral Large 3 | 0.51 | #401 |
| 15 | ERNIE 4.5 300B A47B | 0.5 | #458 |
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-problem-solving · How It Works · Data refreshed daily, snapshot 2026-10-20.