SpecialEduBench - Knowledge: leaderboard
The knowledge axis of SpecialEduBench (Na et al., 2026), a benchmark of pedagogical competence in language intervention for autistic children: four-option questions on medicine, early milieu teaching, special education and psychology, drawn from MedMCQA, MedQA, the Pedagogy Benchmark and MMLU plus 500 new expert-reviewed items. Exact-match accuracy in percent, macro-averaged over the four categories; higher is better.
Metric: Macro-average accuracy over four knowledge categories (%; exact match on four-option items: 3,752 medical items from MedMCQA and MedQA, 500 generated early-milieu-teaching items, 223 special-educational-needs items from the Pedagogy Benchmark and 62 psychology items from MMLU and MMLU-Pro; mean of three runs at temperature 1.0; text only). Source: arxiv.org. Saturation forecast: Estimated already saturated. 8 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 5 (Low) | 93.7 | #20 (Claude Opus 5) |
| 2 | Gemini 3.1 Pro (Preview) | 93.6 | #91 |
| 3 | Grok 4.5 | 92.1 | #78 |
| 4 | Qwen 3 VL 235B A22B (Thinking) | 87.8 | #451 |
| 5 | Mistral Large 3 | 85.2 | #819 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=specialedubench-knowledge · How It Works · Data refreshed daily, snapshot 2026-11-12.