HELM Finance - BANKING77: leaderboard
HELM Finance run of BANKING77, intent classification over 77 fine-grained customer-banking queries (exact match).
Metric: EM. Source: crfm.stanford.edu. Saturation forecast: Estimated already saturated. 22 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4o (2024-05-13) | 78.5 | #325 |
| 2 | Gemini 1.5 Pro (002) | 78.4 | #349 |
| 3 | Claude 3.5 Sonnet (20240620) | 77.7 | #299 |
| 4 | Llama 3.1 405B Instruct | 76.9 | #457 |
| 5 | Claude 3 Opus (20240229) | 75.8 | #412 |
| 6 | Llama 3.2 90B Vision Instruct | 75.8 | #622 |
| 7 | Gemini 1.5 Flash (001) | 74.3 | #577 |
| 8 | Llama 3.1 70B Instruct | 73.9 | #561 |
| 9 | Mixtral 8x22B Instruct | 73.1 | #728 |
| 10 | Qwen 2 72B Instruct | 72.3 | #576 |
| 11 | Llama 3 70B Instruct | 72.2 | #634 |
| 12 | Claude 3 Sonnet (20240229) | 71.3 | #677 |
| 13 | GPT-3.5 Turbo (0125) | 71.3 | #984 |
| 14 | GPT-4o Mini (2024-07-18) | 70.1 | #558 |
| 15 | Qwen 1.5 72B Chat | 68.2 | #776 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=helm-finance-banking77 · How It Works · Data refreshed daily, snapshot 2026-10-20.