openPangu-2.0-Flash: 92B MoE model with 512k context

View on Reddit

The openPangu-2.0-Flash model has 92B total parameters (6B activated) and a 512k context length. It is a MoE model trained on Ascend hardware, but has not yet been uploaded to HuggingFace.