Fin-Bias: leaderboard
Large language models (LLMs) are increasingly deployed in financial contexts, raising critical concerns about reliability, alignment, and susceptibility to adversarial manipulation. While prior finance-related benchmarks assess LLMs' capabilities in stock trading, they are often restricted to small sample and fail to demonstrate LLM susceptibility to context with potential human bias. We introduce
Metric: Average Herding Score (with rating) (self-reported). Source: benchmarklist.com. 19 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3 8B | 98.3 | #674 |
| 2 | GLM-4 9B Chat | 98.1 | #916 |
| 3 | gemma-7B (IT) | 97.6 | #1496 |
| 4 | Gemma 2 27B (IT) | 97.5 | #645 |
| 5 | Claude 3.5 Haiku | 97.1 | #568 |
| 6 | Gemma 2 9B (IT) | 97 | #788 |
| 7 | GPT-4 | 95.9 | #438 |
| 8 | GPT-5 | 94.6 | #98 |
| 9 | Mistral Nemo Instruct (2407) | 90.8 | #936 |
| 10 | Claude Sonnet 4 | 90.6 | #199 |
| 11 | internlm2-chat-7B | 89.1 | #1121 |
| 12 | Yi-1.5-9B-Chat-16K | 83.4 | #1135 |
| 13 | DeepSeek-V2-Lite-Chat | 78 | #1210 |
Interactive version: theaggregate.ai/benchmark?slug=fin-bias · How It Works · Data refreshed daily, snapshot 2026-10-20.