BacktestBench: leaderboard
Quantitative backtesting is essential for evaluating trading strategies but remains hampered by high technical barriers and limited scalability. While Large Language Models (LLMs) offer a transformative path to automate this complex, interdisciplinary workflow through advanced code generation, tool usage, and agentic planning, the practical realization is significantly challenged by the current la
Metric: Overall Accuracy (OA) (self-reported). Source: benchmarklist.com. Saturation forecast: Around December 2026. 23 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Pro | 67.41 | #80 |
| 2 | Qwen 3 Max | 61.84 | #208 |
| 3 | Seed 1.8 | 60.58 | #139 |
| 4 | GLM-4.7 | 56.83 | #192 |
| 5 | Qwen 3 235B A22B | 55.01 | #313 |
| 6 | DeepSeek V3.2 | 50.7 | #205 |
| 7 | GPT-OSS-120B | 49.56 | #341 |
| 8 | Qwen 3 32B | 47.81 | #443 |
| 9 | Qwen 3 30B A3B | 47.03 | #493 |
| 10 | MiniMax-M2.1 | 45.91 | #289 |
| 11 | Kimi K2 (Thinking) | 44.97 | #243 (Kimi K2) |
| 12 | Qwen 3 Next 80B A3B | 44.06 | #320 |
| 13 | Qwen 3 Coder Plus | 42.05 | #338 |
| 14 | MiMo-V2-Flash | 40.43 | #372 |
| 15 | Qwen 3 14B | 36.97 | #537 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=backtestbench · How It Works · Data refreshed daily, snapshot 2026-10-20.