BizBench - ConvFinQA Extract: leaderboard
Metric: Accuracy (%; 916 ConvFinQA questions answerable by extracting a number from the report text or table, 3-shot). Source: arxiv.org. Saturation forecast: Estimated already saturated. 16 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-4 | 94 |
| 2 | Llama 2 70B | 92.8 |
| 3 | Mixtral 8x7B | 92.8 |
| 4 | GPT-3.5 | 92.4 |
| 5 | Mistral 7B | 91.5 |
| 6 | Llama 2 13B | 88.2 |
| 7 | Llama 2 7B | 86.1 |
| 8 | mpt-30B | 85.5 |
| 9 | falcon-40B | 82.8 |
| 10 | starcoder | 79.7 |
| 11 | mpt-7B | 71.7 |
| 12 | falcon-7B | 66.5 |
Interactive version: theaggregate.ai/benchmark?slug=bizbench-convfinqa-extract · How It Works · Data refreshed daily, snapshot 2026-09-26.