MiniMax H3 BF16 generator subset

This is a slim, unquantized BF16 generator cache derived from MiniMaxAI/MiniMax-H3 for a split RunPod worker. The model-card text has been modified to describe this subset.

It contains transformer, vae, audio_vae, scheduler, audio_scheduler, and modular_model_index.json. The external H3 conditioner is intentionally not included and is called through its existing Hugging Face Space API.

The weight and configuration files were copied server-side from the official repository without conversion or quantization. Use ovedrive/MiniMax-H3-generator-bf16 in the RunPod endpoint Model field; no Hugging Face model-cache token is required.

See the official model repository for full documentation: https://huggingface.co/MiniMaxAI/MiniMax-H3

Downloads last month
-
Safetensors
Model size
33B params
Tensor type
F32
·
BF16
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for ovedrive/MiniMax-H3-generator-bf16

Finetuned
(16)
this model