LogiQAte - Obfus Direction Sense: leaderboard

Metric: Exact-match accuracy (%; direction puzzles padded with self-cancelling detours, zero-shot, temperature 0). Source: arxiv.org. Saturation forecast: Around December 2026. 10 models tracked.

Top models

#ModelScore
1O4 Mini45
2GPT-544
3Gemini 2.5 Pro41
4QwQ-32B37
5Gemma 3 27B (IT)17
6Claude 3.7 Sonnet17
7GPT-4o14
8Claude Haiku 4.510
9GPT-4o Mini5

Interactive version: theaggregate.ai/benchmark?slug=logiqate-obfus-direction-sense · How It Works · Data refreshed daily, snapshot 2026-09-26.