FORTIS - Skill Selection Fail Rate: leaderboard
Metric: Fail rate (%), the share of over-privileged or no-action outcomes, on FORTIS Task 1 (skill selection, 600 queries over email, e-commerce and filesystem libraries of 20 skills each, privilege levels hidden), temperature 0; lower is better. Source: arxiv.org. Saturation forecast: Not forecast. 10 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Opus 4.7 | 35.5 |
| 2 | Qwen 3.6 Max | 38.5 |
| 3 | Claude Sonnet 4.6 | 40.3 |
| 4 | Gemini 3.1 Pro (Preview) | 41.2 |
| 5 | Gemini 3 Flash | 43.7 |
| 6 | DeepSeek V4 Flash | 44.5 |
| 7 | GPT-5.4 Mini | 45.5 |
| 8 | Kimi K2.6 | 47.8 |
| 9 | GPT-5.5 | 51.2 |
| 10 | GPT-5.4 | 52.7 |
Interactive version: theaggregate.ai/benchmark?slug=fortis-skill-selection-fail-rate · How It Works · Data refreshed daily, snapshot 2026-10-07.