Grok 3 Beta – The Age of Reasoning Agents

You can try it at https://grok.com/ in case it wasn't obvious.

These new reasoning models: Deepseek R1, Chat-GPT (whatever it's called), Grok v3 – they're qualitatively different to what I picked up a year-or-so ago and put down again due to hitting their limitations pretty quick.

As of Feb '25 I can say that I can see sparks of AGI in the current crop of reasoners. Pair these models with a robot body that interacts with the world so that the AI can link up sensorimotor data with its language model and we'll have walking-talking smart enough robots within a decade. We're going to have actually useful smart AI assistants on our mobile devices within a couple of years, and they're going to be run locally in a useful manner on device for all personal needs when mobile device memory hits, what, 64gb?