Scale AI - TutorBench: leaderboard
Evaluates AI performance on high school and AP-level tutoring tasks. Tests pedagogical capability and educational content quality.
Metric: Score. Source: scale.com. Status: saturation imminent. 27 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Muse Spark | 68.55 |
| 2 | GPT-5.4 Pro (xHigh) | 56.62 |
| 3 | Gemini 2.5 Pro (Preview 06-05) | 55.65 |
| 4 | GPT-5 | 55.33 |
| 5 | Kimi K2.5 | 54.56 |
| 6 | GPT-5.1 (Thinking) | 54.09 |
| 7 | Claude Opus 4.6 (Adaptive Reasoning, Max Effort) | 53.68 |
| 8 | Gemini 3 Pro (Preview) | 53.67 |
| 9 | Claude Opus 4.6 (Non-reasoning) | 53.55 |
| 10 | GPT-5.2 | 53.49 |
| 11 | Gemini 3.1 Pro (Preview) | 52.99 |
| 12 | O3 (2025-04-16) (Medium) | 52.76 |
| 13 | O3 (2025-04-16) (High) | 52.09 |
| 14 | Gemini 3.1 Flash Lite (Preview) | 51.5 |
| 15 | Claude Opus 4.5 (20251101) (Thinking) | 51.2 |
Interactive version: theaggregate.ai/benchmark?slug=scale-ai-tutorbench · How It Works · Data refreshed daily, snapshot 2026-09-05.