CorpusQA (128K) - Education: leaderboard

Metric: Accuracy (%; 82 questions over English education documents, 128K-token corpora, DeepSeek-V3 judge). Source: arxiv.org. Saturation forecast: Estimated already saturated. 8 models tracked.

Top models

#ModelScore
1GPT-581.71
2DeepSeek R1 052878.05
3Gemini 2.5 Pro76.83
4MiniMax M1 80k76.83
5Gemini 2.5 Flash74.39
6GPT-5 Mini74.39
7Qwen 3 235B A22B 2507 (Thinking)70.73
8Qwen 3 30B A3B 2507 (Thinking)64.63

Interactive version: theaggregate.ai/benchmark?slug=corpusqa-128k-education · How It Works · Data refreshed daily, snapshot 2026-09-25.