StudentBench - Practice Design: leaderboard
Metric: Expert tutor pairwise preference, Bradley-Terry score (practice problems). Source: studentbench.org. Saturation forecast: Around December 2027. 13 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Opus 5 (High) | 1.46 |
| 2 | Claude Opus 4.8 (xHigh) | 1.14 |
| 3 | Claude Opus 4.8 (Non-reasoning) | 0.96 |
| 4 | GPT-5.5 (High) | 0.87 |
| 5 | Claude Sonnet 4.6 (Low) | -0.04 |
| 6 | Claude Sonnet 5 (Low) | -0.26 |
| 7 | Gemini 3.1 Pro (Preview) (High) | -0.55 |
| 8 | Gemini 3.7 Flash (Medium) | -0.62 |
| 9 | Gemini 3.5 Flash (Low) | -1.3 |
| 10 | GPT-5.4 Mini (Non-reasoning) | -1.31 |
| 11 | Gemma 4 31B (High) | -1.4 |
Interactive version: theaggregate.ai/benchmark?slug=studentbench-practice-design · How It Works · Data refreshed daily, snapshot 2026-10-04.