Open FinLLM Reasoning - DM-Complong: leaderboard
Open FinLLM Reasoning subtask on long, complex financial documents requiring multi-step numeric and contextual reasoning.
Metric: Accuracy (%). Source: huggingface.co. Saturation forecast: Estimated already saturated. 26 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | DeepSeek V3 | 42.33 | #321 |
| 2 | GPT-4o | 39.33 | #343 |
| 3 | GPT-4.5 | 39.33 | #274 |
| 4 | DeepSeek R1 | 38.67 | #251 |
| 5 | O1 | 36.67 | #183 |
| 6 | Llama 3.1 70B Instruct | 34.33 | #561 |
| 7 | Llama 3.3 70B Instruct | 32 | #534 |
| 8 | DeepSeek R1 Distill Llama 70B | 30.67 | #594 |
| 9 | Qwen 2.5 32B Instruct | 30 | #498 |
| 10 | Qwen 2.5 14B Instruct | 26.67 | #647 |
| 11 | DeepSeek R1 Distill Qwen 32B | 24.67 | #655 |
| 12 | DeepSeek R1 Distill Qwen 14B | 21 | #844 |
| 13 | QwQ-32B | 20 | #421 |
| 14 | Qwen 2.5 7B Instruct | 17.67 | #862 |
| 15 | DeepSeek R1 Distill Llama 8B | 15.67 | #1291 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=open-finllm-reasoning-dm-complong · How It Works · Data refreshed daily, snapshot 2026-10-20.