LongCat 2.0: Open Mixture-of-Experts Model with 1.6T Parameters

benjiro29 released LongCat 2.0, a large open-source MoE model with 1.6 trillion total parameters but only 48 billion active during inference. For indie developers building AI applications, MoE architectures like this offer a path to competitive performance without the compute cost of dense models—useful if you're running inference on tight margins.