AAGI Read More Even the latest AI models make three systematic reasoning errors, ARC-AGI-3 analysis showsMay 2, 2026 No frontier model cracks the 1 percent mark on the ARC-AGI-3 leaderboard. GPT-5.5 leads with 0.4 percent at…