Not too dissimilar to the first HC1 (6nm 815mm² 53B Transistors embedding an 8b LLM):

> Our second model, still based on Taalas’ first-generation silicon platform (HC1), will be a mid-sized reasoning LLM