BacktestBench - Strategy Selection: leaderboard
Metric: Back-test accuracy (%) on the Strategy Selection task family (choosing which of several strategies meets a stated outcome), synthetic BacktestBench test split (the expert-crafted subset excluded), each model driving the AutoBacktest pipeline of the authors (Summarizer, SQL Retriever and Python Coder agents) over a PostgreSQL market database, temperature 0.6; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 23 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3 Pro | 89.07 |
| 2 | Qwen 3 Max | 85.27 |
| 3 | Qwen 3 235B A22B 2507 (Thinking) | 85.04 |
| 4 | Seed 1.8 | 84.8 |
| 5 | DeepSeek V3.2 | 81.47 |
| 6 | GLM-4.7 | 80.76 |
| 7 | Qwen 3 Coder Plus | 77.91 |
| 8 | MiniMax-M2.1 | 76.72 |
| 9 | GPT-OSS-120B | 76.48 |
| 10 | Qwen 3 30B A3B 2507 (Thinking) | 75.06 |
| 11 | Qwen 3 32B | 72.21 |
| 12 | MiMo-V2-Flash | 71.26 |
| 13 | Qwen 3 Next 80B A3B (Thinking) | 68.17 |
| 14 | Kimi K2 (Thinking) | 67.46 |
| 15 | Qwen 3 14B | 64.61 |
Interactive version: theaggregate.ai/benchmark?slug=backtestbench-strategy-selection · How It Works · Data refreshed daily, snapshot 2026-10-07.