CorpusQA (128K) - Financial (English): leaderboard
English financial-filing domain of CorpusQA (Lu et al., Tongyi Lab, 2026) at the 128K-token scale (concatenated corpora of about 92K tokens): 83 analytical questions (filters, rankings, counts and computed values) answered over the whole concatenated corpus of English financial filings, with reference answers computed by SQL over fields extracted from every document. Accuracy in percent as judged by DeepSeek-V3 against the reference; higher is better.
Metric: Accuracy (%; 83 questions over English financial filings, 128K-token corpora, DeepSeek-V3 judge). Source: arxiv.org. Saturation forecast: Estimated already saturated. 8 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 2.5 Pro | 80.72 | #262 |
| 2 | GPT-5 | 78.31 | #163 |
| 3 | GPT-5 Mini (2025-08-07) | 78.31 | #341 |
| 4 | Gemini 2.5 Flash | 74.7 | #452 |
| 5 | Qwen 3 235B A22B 2507 (Thinking) | 74.7 | #460 |
| 6 | DeepSeek R1 0528 | 73.49 | #425 |
| 7 | MiniMax M1 80k | 67.07 | #926 |
| 8 | Qwen 3 30B A3B 2507 (Thinking) | 61.45 | #759 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=corpusqa-128k-financial-english · How It Works · Data refreshed daily, snapshot 2026-11-12.