Tag
A municipal employee in Brazil claims to have discovered a method that makes LLM fine-tuning 1000x faster, though analysis suggests the resulting model, Rio 3.5, is essentially a mixture of existing open-source models Nex N2 Pro and Qwen 3.5.