Expired50% confidenceFactExact time
Starting from an update approximately three weeks ago, the MiniMax H3 video generation model has been running approximately 4 times slower at full resolution.
1
Sources
50%
Confidence
Short-term (~14 days)
Relevance
8/30/2026
First Seen
This claim has expired. It is kept for the record and no longer describes the current state.
Valid until: 9/13/2026(expired)
Sources
Related Entities
Related Claims
UnverifiedMiniMax H3 supports up to 2K resolution, with individual video clips up to 15 seconds in length.74% similarUnverifiedWhen processing every frame of a 30fps video with a VLM, the processing speed of a pure VLM solution may be 10 to 100 times slower than real-time requirements.67% similarUnverifiedThe starting point of the H3 video generation speed regression has been pinpointed to ComfyUI version v0.32.0, with the related issue recorded in GitHub Issue #15665.67% similarUnverifiedAccessing the H3 model through the official app MiniMax Design costs approximately 6 cents per second at 2K resolution.65% similarUnverifiedRunning MiniMax H3 on an RTX 4070 or below (12 GB or less) may require extreme optimizations such as full model offloading to CPU, reducing resolution, and decreasing frame count.63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/825613API
curl https://kongchang.com/api/v1/knowledge/claims/825613MCP
get_claim(id=825613)