@Walker notable comment, i mean obligatory comment - likely based on wireguard which is a lot faster than openvpn last time i looked. This could be for better security stance and attribution #lez zepplin #source wsj #fair use #elo is ranking #intelligence per gigabyte quotient #sumdensity moved #ai pagerank
per ai itself these are stats - may or may not be accurate - these things are on edge of being smarter than us, maybe that is what that article is intimating (doh)
As of early 2026, the best large language model (LLM) AI models are performing within the
120 to 140+ IQ range (genius level) based on specialized, adapted cognitive benchmarks.
The top-performing models in early 2026, such as OpenAI's o3 and Claude 4.5, have demonstrated capabilities far exceeding the average human IQ (90–110), with some estimates suggesting frontier models are reaching an "human IQ-equivalent" of 150-160, according to data from Tracking AI and analyses by experts.
Based on data from 2025–2026 and benchmarks that adapt Mensa-style tests for LLMs, the estimated IQ scores of top AI models are:
OpenAI o3/o3-mini: ~135–136 (Genius Range)
Anthropic Claude 4.5/4 Sonnet: ~127–130
Google Gemini 2.0/2.5 Pro: ~124–130
Grok-3/4: ~112–126
DeepSeek R1: ~106–110
Key trends include:
Rapid Progression: Top AI models increased from an estimated 96 IQ to over 135 IQ in one year (2024–2025).