INS-ActBench - Practice: leaderboard

Metric: Numerical accuracy (%; 0.2 percent relative tolerance) on the 1,152 spreadsheet and R-code tasks (INS-Act-Practice); 12,050 questions from the exams of 16 actuarial associations; knowledge and case questions two-shot, practice tasks zero-shot; proprietary models with thinking enabled at high where supported. Source: arxiv.org. Saturation forecast: Around December 2026. 9 models tracked.

Top models

#ModelScore
1GPT-5.5 (High)69.2
2Gemini 3.1 Pro (Preview)67.9
3Claude Opus 4.7 (High)65.4
4DeepSeek V4 Pro (High)61.7
5Qwen 3.6 Plus57.4
6Kimi K2.647.2
7Qwen 3.5 35B A3B37.7
8Qwen 3 14B26
9Gemma 3 12B (IT)20.4

Interactive version: theaggregate.ai/benchmark?slug=ins-actbench-practice · How It Works · Data refreshed daily, snapshot 2026-09-29.