Benchmarks
New Benchmark Results: iCon Modular AI Hits Top Scores Paid Members Public
humanity.ai's modular AI beats leading models on key benchmarks, delivering superior reasoning with 300B parameters, all running on MacBooks, not datacenters. Hardware-efficient, private, and fully scalable.
ARC-AGI-2: Challenge Accepted Paid Members Public
humanity.ai is gearing up for ARC-AGI-2, one of the world’s leading AGI benchmarks. Our team is tuning our modular system to take on the challenge—and prove what a self-evolving architecture can really do.