SciIBI (Vision+Text, CoT): leaderboard
Metric: Accuracy (%) of 4-way Core Instructional Practices classification (big idea, eliciting ideas, guiding sense-making, pressing for evidence) of 113 K-12 science classroom clips (82 NGSS-aligned lessons), expert consensus labels, vision and text input (transcript plus up to 30 frames sampled at 1 fps), step-by-step reasoning before the label, temperature 0; accuracy counts parsed JSON answers; higher is better. Source: arxiv.org. Saturation forecast: Around June 2027. 4 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4o | 51.9 | #333 |
| 2 | Qwen 3 VL 235B A22B | 49.1 | #228 |
| 3 | Gemini 2.5 Pro | 48.2 | #145 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=sciibi-vision-plustext-cot · How It Works · Data refreshed daily, snapshot 2026-10-11.