CSTutorBench: leaderboard
CSTutorBench (2026) evaluates language models as tutors for middle-school block-based programming in VEX VR: 17 scenario questions with Blockly code snapshots, scored 0 to 2 per rubric criterion by an LLM judge with human review of every judgment. The overall score is the percent of rubric points earned; higher is better.
Metric: Overall rubric score (%; points earned out of the maximum on 17 scenario-based VEX VR block-programming tutoring questions for middle-school students, each scored 0 to 2 per criterion by a Claude Sonnet judge with human review of every judgment; revised system prompt (Trial 2); all eight criteria per question; higher is better). Source: arxiv.org. Saturation forecast: Around February 2027. 11 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemma 4 31B (IT) | 89 | #342 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=cstutorbench · How It Works · Data refreshed daily, snapshot 2026-11-13.