FinSkillBench (No-Skill): leaderboard
Metric: Task score (0-1, native; unweighted mean subtask score with no skill corpus or finance tools available; over FinSkillBench's 12 investment-management subtasks (portfolio construction, risk management, fundamental analysis), unweighted mean, hidden ground truth and task-specific verifiers). Source: arxiv.org. Saturation forecast: Around December 2026. 9 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Sonnet 4.6 | 0.53 |
| 2 | DeepSeek V3.2 | 0.46 |
| 3 | GPT-5.4 | 0.43 |
| 4 | Gemini 3.1 Flash Lite | 0.42 |
| 5 | Gemini 2.5 Pro | 0.38 |
| 6 | GPT-4.1 | 0.33 |
| 7 | GLM-5.1 | 0.27 |
| 8 | Grok 4 | 0.1 |
| 9 | Phi-4 | 0 |
Interactive version: theaggregate.ai/benchmark?slug=finskillbench-no-skill · How It Works · Data refreshed daily, snapshot 2026-09-26.