Does Qubic outperform Grok on an AI benchmark?
Qubic claims its Neuraxon 2.0 architecture achieved 0.13% on the ARC-AGI-3 benchmark, compared to 0.10% for Grok 4.2 beta.
ARC-AGI-3 is one of the toughest and most respected benchmarks for measuring artificial general intelligence (AGI) of AIs.
Unlike classic benchmarks where AIs can “cheat” by memorizing data, ARC-AGI tests the ability to reason on new tasks never seen during training.
Version 3 (ARC-AGI-3) is even harder because it is interactive. The AI must explore an environment, understand the rules by itself, adapt and solve problems as a human would.
Human score ≈ 100 %
For fun I’ll give you a short list of top scores:
Anthropic Opus 4.8 : ~1.5%
Gemini 3.1 Pro : ~0.4%
GPT-5.5 : ~0.4%
Neuraxon 2.0 (Qubic): 0.13%
Grok 4.20 : 0.1%
For those who don’t quite understand, there are two doctors working on an AI inspired by how our brain works that will later be deployed on the Qubic network. As you can see, current AIs are just huge databases and none reason on their own; we want an AI that reasons by itself, which we call AGI. In this arena, Qubic is battling companies worth hundreds of billions.
I’ll say it again, Qubic is not for the impatient. It is even quite inflationary at the moment. But if their tech works, the market won’t be ready. You have to know we are talking about technology, so no one is certain; it could be that tomorrow we learn a new project from Asia has succeeded at what they’re trying to do. I’m emphasizing that people stay measured in their investment and don’t fall in love. I’m talking about tech, not price.
