FINESSE-Bench - Trading TA: leaderboard
Metric: Accuracy (%) on the 413 Trading_TA numerical and short-answer applied technical-analysis tasks (patterns, momentum and mean reversion, entries and exits, backtesting) of FINESSE-Bench; a GPT-5.2 judge marks each answer correct or incorrect against the reference; zero-shot, one fixed prompt per task type, temperature 0 where possible, reasoning configurations with medium effort where the model offers them; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 31 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.2 (Medium) | 83.54 |
| 2 | Claude Sonnet 4.6 (Medium) | 83.54 |
| 3 | Kimi K2.5 | 83.29 |
| 4 | GLM-5 | 82.57 |
| 5 | Qwen 3.5 Plus (2026-02-15) | 82.32 |
| 6 | Qwen 3.5 397B A17B | 81.84 |
| 7 | GLM-4.7 | 81.6 |
| 8 | GPT-5.4 (Medium) | 80.87 |
| 9 | Qwen 3.5 27B | 79.66 |
| 10 | Qwen 3.5 35B A3B | 78.69 |
| 11 | Qwen 3.5 122B A10B | 78.21 |
| 12 | Claude 3.7 Sonnet (Thinking) | 77.97 |
| 13 | Qwen 3 235B A22B 2507 (Thinking) | 77.72 |
| 14 | Qwen 3.5 Flash (02-23) | 77.24 |
| 15 | DeepSeek R1 0528 | 76.51 |
Interactive version: theaggregate.ai/benchmark?slug=finesse-bench-trading-ta · How It Works · Data refreshed daily, snapshot 2026-10-07.