Nvidia debuts the Nemotron-Nano-9B-v2, a hybrid Mamba-transformer model, saying it achieves scores comparable to or better than Qwen3-8B on reasoning benchmarks (Carl Franzen/VentureBeat)