MiniMax H3 Video Model Going Open-Weight: 33B Main DiT & 20B Pruned Variant
EverythingMacPro · reddit · 2026-08-02
Based on open PRs for ComfyUI and Hugging Face diffusers, the MiniMax H3 video generation model will be open-weight in under 6 hours.
Key specifications revealed:
- 33B parameters for the main DiT architecture, alongside a pruned 20B variant.
- Utilizes Qwen-3-VL-32b as the text encoder.
More from Models
- ChatGPT Zero-Shots a 15-Year-Old Open Problem in Information Theory — DimitrisPapail · 2026-08-02
- Report: OpenAI's Internal Astra Model Solved 10 Major Math and CS Problems — wa5ina · 2026-08-02
- Dev Finds LLM Actively Trying to Game Benchmarks to Disprove Results — wavefnx · 2026-08-02
- Scaling Inference Compute is the Next Frontier for AI Models — JFPuget · 2026-08-02
- LLMs Decode Early MDLM Gibberish: Predict The Verge as Training Data Source — dejanseo · 2026-08-02
- Anthropic Employee Replicates Half of Astra Proofs Using Fable, Sparking Debate — Outside-Iron-8242 · 2026-08-02