FinED-Bench (English): leaderboard
FinED-Bench-EN (2026) is the English companion set of FinED-Bench: 56 English-language financial documents with injected general knowledge, domain knowledge and reasoning errors that a model must locate and classify. F1 in percent: a detection counts when the extracted sentence contains the erroneous one and the error type is correct; higher is better.
Metric: F1 (%; FinED-Bench-EN, 56 English-language financial documents with annotated general-knowledge, domain-knowledge and reasoning errors; a predicted error counts as correct when the extracted sentence matches or contains the annotated sentence and the error type is classified correctly; the paper's error-detection prompt over each document, reasoning models with thinking on unless marked no thinking). Source: arxiv.org. Saturation forecast: Around February 2027. 9 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5 | 53.38 | #163 |
| 2 | Qwen 3 14B | 44.52 | #1235 |
| 3 | Qwen 3 8B | 38.64 | #1838 |
| 4 | DeepSeek R1 0528 Qwen3 8B | 33.94 | #2101 |
| 5 | Qwen 3 14B (Non-reasoning) | 23.6 | #1235 (Qwen 3 14B) |
| 6 | GPT-4o Mini | 21.2 | #1627 |
| 7 | Qwen 3 8B (Non-reasoning) | 17.58 | #1838 (Qwen 3 8B) |
| 8 | Llama 3.1 8B | 12.9 | #4900 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=fined-bench-english · How It Works · Data refreshed daily, snapshot 2026-11-12.