非自研模型的量化轉存,方便地端部署。Quantized re-uploads of third-party models for on-premise deployment.
Note NVIDIA Nemotron 3 Nano 的 GGUF