Depends on what weights, I landed on:
diffusion model: minimax_h3_fl2va_bf16.safetensors text encoder: qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors video VAE: minimax_h3_video_vae_fp16.safetensors audio VAE: minimax_h3_audio_vae_fp32.safetensors