FINESSE-Bench - CMT-like Level 2: leaderboard
Metric: Accuracy (%) on the 251 CMT-like Level 2 multiple-choice questions on technical analysis theory, chart patterns, indicators and trading-system testing 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: Estimated already saturated. 31 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GLM-5 | 89.64 |
| 2 | Claude Sonnet 4.6 (Medium) | 89.24 |
| 3 | Qwen 3.5 Plus (2026-02-15) | 89.24 |
| 4 | Qwen 3.5 397B A17B | 88.84 |
| 5 | Kimi K2.5 | 88.05 |
| 6 | GLM-4.7 | 88.05 |
| 7 | Qwen 3.5 27B | 86.45 |
| 8 | Qwen 3.5 122B A10B | 86.06 |
| 9 | DeepSeek R1 0528 | 85.66 |
| 10 | GPT-5.4 (Medium) | 85.66 |
| 11 | GPT-5.2 (Medium) | 85.26 |
| 12 | Qwen 3 235B A22B 2507 (Thinking) | 84.46 |
| 13 | MiniMax-M2.1 | 84.46 |
| 14 | MiniMax-M2.5 | 84.06 |
| 15 | Qwen 3.5 35B A3B | 83.67 |
Interactive version: theaggregate.ai/benchmark?slug=finesse-bench-cmt-like-level-2 · How It Works · Data refreshed daily, snapshot 2026-10-07.