SupraBench - Top-Binder Selection: leaderboard
Metric: Accuracy (%) at picking the strongest of four candidate guests for a host (2,264 questions), zero-shot direct prompt. Source: arxiv.org. Saturation forecast: Around April 2028. 8 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3 Flash | 49.8 |
| 2 | GPT-5.4 Mini | 42.8 |
| 3 | GPT-5.4 Nano | 41.1 |
| 4 | Qwen 3.5 27B | 40.4 |
| 5 | Qwen 3.5 9B | 37.9 |
| 6 | Llama 3.1 70B Instruct | 33.8 |
| 7 | Llama 3.1 8B Instruct | 22.8 |
Interactive version: theaggregate.ai/benchmark?slug=suprabench-top-binder-selection · How It Works · Data refreshed daily, snapshot 2026-09-29.