BacktestBench - Parameter Confirmation: leaderboard
Metric: Back-test accuracy (%) on the Parameter Confirmation task family (confirming the strategy parameter that produces 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 | 82.46 |
| 2 | Qwen 3 Max | 82.28 |
| 3 | GLM-4.7 | 77.24 |
| 4 | Seed 1.8 | 77.05 |
| 5 | Qwen 3 235B A22B 2507 (Thinking) | 75.93 |
| 6 | DeepSeek V3.2 | 73.51 |
| 7 | Qwen 3 Coder Plus | 71.08 |
| 8 | GPT-OSS-120B | 70.34 |
| 9 | Qwen 3 32B | 70.34 |
| 10 | MiniMax-M2.1 | 69.22 |
| 11 | Qwen 3 Next 80B A3B (Thinking) | 65.86 |
| 12 | Qwen 3 30B A3B 2507 (Thinking) | 65.86 |
| 13 | MiMo-V2-Flash | 64.37 |
| 14 | Qwen 3 14B | 61.57 |
| 15 | Kimi K2 (Thinking) | 58.02 |
Interactive version: theaggregate.ai/benchmark?slug=backtestbench-parameter-confirmation · How It Works · Data refreshed daily, snapshot 2026-10-07.