OpenLearnLM - Skills - Teaching: leaderboard
Teaching tasks from the OpenLearnLM Benchmark (Lee et al., 2026): generated scenarios for instructional designers, content creators and teaching assistants, such as curriculum design, learner analysis and error detection, mostly answered in long form. An LLM judge, not named in the paper, scores each answer on a scenario-specific 1-10 rubric; the score is the mean, and higher is better.
Metric: Rubric Score (1-10; LLM judge, scenario-specific rubrics, Teaching center). Source: arxiv.org. Saturation forecast: Estimated already saturated. 7 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 4.5 | 8.83 | #152 |
| 2 | GPT-5.2 | 8.8 | #172 |
| 3 | Kimi K2 (Thinking) | 8.68 | #435 (Kimi K2) |
| 4 | DeepSeek V3.2 | 8.53 | #403 |
| 5 | Grok 4.1 Fast | 8.44 | #426 |
| 6 | Gemini 3 Pro | 8.26 | #135 |
| 7 | GLM-4.7 | 8.16 | #365 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=openlearnlm-skills-teaching · How It Works · Data refreshed daily, snapshot 2026-11-12.