Open FinLLM Reasoning - DM-Complong: leaderboard
Metric: Accuracy (%). Source: huggingface.co. 26 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | DeepSeek V3 | 42.33 |
| 2 | GPT-4o | 39.33 |
| 3 | GPT-4.5 | 39.33 |
| 4 | DeepSeek R1 | 38.67 |
| 5 | O1 | 36.67 |
| 6 | Llama 3.1 70B Instruct | 34.33 |
| 7 | Llama 3.3 70B Instruct | 32 |
| 8 | DeepSeek R1 Distill Llama 70B | 30.67 |
| 9 | Qwen 2.5 32B Instruct | 30 |
| 10 | Qwen 2.5 14B Instruct | 26.67 |
| 11 | DeepSeek R1 Distill Qwen 32B | 24.67 |
| 12 | DeepSeek R1 Distill Qwen 14B | 21 |
| 13 | QwQ-32B | 20 |
| 14 | Qwen 2.5 7B Instruct | 17.67 |
| 15 | DeepSeek R1 Distill Llama 8B | 15.67 |
Interactive version: theaggregate.ai/benchmark?slug=open-finllm-reasoning-dm-complong · How It Works · Data refreshed daily, snapshot 2026-09-05.