Unverified50% confidenceFactExact time
The updated VAE model filename is minimax_h3_video_vae_int8_convrot.safetensors.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
9/18/2026
First Seen
Valid until: 12/17/2026
Sources
Related Entities
Related Claims
UnverifiedKijai updated the VAE model for MiniMax-H3 (MH3), adopting an int8 quantized version that significantly reduces VRAM usage with almost no loss in image quality.65% similarUnverifiedModel loading is divided into three categories: Clip model (select the int8 version), UNet large model (minimax_h3_int8 version), and VAE decoder.61% similarUnverifiedvLLM v0.27.0 adds support for the Qwen3.5 series (including dense and MoE models) and supports EVS video token pruning.57% similarUnverifiedThis practical test deployed DeepSeek-V4-Flash (DSpark version) via GPUStack in an 8-GPU H20-141G environment.56% similarUnverifiedModels newly supported in SGLang v0.5.18 include Muse Glimmer, Intern-S2-Mobius, SANA-Video, LingBot-Video-MoE, LTX-2.5, Cosmos3 Edge & Distilled, and LongCat-Image.55% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/931903API
curl https://kongchang.com/api/v1/knowledge/claims/931903MCP
get_claim(id=931903)