CorpusQA (1M): leaderboard
Corpus-level analysis benchmark (CorpusQA, Lu et al., Tongyi Lab, 2026) at its 1M-token scale: 329 analytical questions over concatenated real-document corpora of about 905K tokens, run only on models that accept million-token input, with reference answers computed by SQL over fields extracted from every document. A DeepSeek-V3 judge checks semantic equivalence with the reference; the score is the mean of the four domain accuracies in percent, and higher is better.
Metric: Accuracy (%; mean of the four domain accuracies, 1M-token corpora, DeepSeek-V3 judge). Source: arxiv.org. Saturation forecast: Estimated already saturated. 3 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 2.5 Pro | 50.73 | #152 |
| 2 | Gemini 2.5 Flash | 37.31 | #245 |
| 3 | MiniMax M1 80k | 11.63 | #432 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=corpusqa-1m · How It Works · Data refreshed daily, snapshot 2026-10-20.