Based on the original Minimax H3 BF16 model, it was merged with the LoRA at a weight of 1.0. ComfyOrg Model Tool was used. To make it exactly identical to the official Minimax H3 model, Block 0 and Block 1 were not quantized. The file size and quantization structure are identical to the official model.
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support