The launch of Deepseek-R1 took the Gen AI industry by storm. It comes a few months after Open AI’s io1 was launched, but is much more economical with its use of the CPU and GPU. Let’s explore how powerful the reasoning capabilities of these models are. But before we do that, it’s useful to spend some time thinking about what reasoning even is. LLMs are designed very differently from human brains, so it would be futile to compare the specific mechanics of how we reason vs how an LLM might do so. However, there are certain basic characteristics that we can fairly expect from anything that claims to reason. ...