New AI model processes massive documents faster with less compute

Original: SubQ: a sub-quadratic LLM with 12M-token context

Source: subq.ai ↗