BizBench - FinCode: leaderboard

Metric: Accuracy (%; 121 CFA and CPA exam questions answered by generating Python code, answer within 1% of the reference, 8-shot). Source: arxiv.org. Saturation forecast: Estimated already saturated. 16 models tracked.

Top models

#ModelScore
1GPT-463.6
2GPT-3.536.1
3Llama 2 70B24
4Mixtral 8x7B19
5Mistral 7B15.7
6starcoder9.9
7falcon-40B8.3
8Llama 2 7B7.4
9Llama 2 13B6.6
10mpt-30B6.6
11mpt-7B6.6
12falcon-7B3.3

Interactive version: theaggregate.ai/benchmark?slug=bizbench-fincode · How It Works · Data refreshed daily, snapshot 2026-09-26.