FinTradeBench (RAG): leaderboard
Metric: Normalized correctness: a calibrated Claude Sonnet 4.5 judge scores each answer against the expert-validated gold answer and golden indicators on a 1 to 5 scale, averaged and divided by 5, times 100 (20 to 100), over all of FinTradeBench's 1,400 questions on NASDAQ-100 companies over 2015 to 2025 (fundamentals, trading-signal and hybrid categories, weighted by their question counts); with dual-track retrieval over SEC filings and daily price data, each model self-selecting its best answer across the retrieval TELeR prompt levels L5 and L6; higher is better. Source: arxiv.org. Saturation forecast: Around 2031. 14 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | DeepSeek R1 | 37.7 | #245 |
| 2 | GPT-5 Mini | 36.4 | #176 |
| 3 | Qwen 2.5 32B Instruct | 36.2 | #491 |
| 4 | DeepSeek R1 Distill Qwen 32B | 33.9 | #640 |
| 5 | Gemini 2.5 Flash | 33.4 | #237 |
| 6 | DeepSeek R1 Distill Qwen 14B | 32.5 | #828 |
| 7 | Gemini 2.5 Flash Lite | 31 | #413 |
| 8 | Phi-4 | 30.8 | #701 |
| 9 | Llama 3.1 8B Instruct | 28.4 | #1018 |
| 10 | Llama 3.3 70B Instruct | 28.4 | #520 |
| 11 | DeepSeek R1 Distill Llama 70B | 26.3 | #582 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=fintradebench-rag · How It Works · Data refreshed daily, snapshot 2026-10-11.