Unverified50% confidenceBenchmarkExact time
First-frame rendering for the MiniMax-H3 model can complete up to approximately one minute earlier, and peak VRAM usage on A100, B200, and RTX PRO 6000 has been reduced by 25–29 GiB.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
9/28/2026
First Seen
Valid until: 12/27/2026
Sources
Related Entities
Related Claims
UnverifiedMiniMax H3 REF completes native 4096×1536 resolution image generation in approximately 1 minute on a single RTX 5060 Ti 16GB GPU with 18 sampling steps.75% 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.74% similarUnverifiedOn an NVIDIA RTX 5090 graphics card, completing one generation with the full MiniMax H3 fusion workflow takes approximately 198.40 seconds.71% similarUnverifiedThe MiniMax H3 standard workflow requires a minimum of 16GB VRAM to run, while 6GB graphics cards can also participate in partial workflows through optimizations such as FP8, INT4 quantization, and chunked inference Tiling.68% similarUnverifiedWhen VRAM is sufficient and the model is fully resident on the GPU, an RTX 4090 running a Q4-quantized 27B model can typically achieve 15–30 tokens/s; once CPU offloading is triggered, the speed may drop sharply to 1–2 tokens/s or even lower.66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/956830API
curl https://kongchang.com/api/v1/knowledge/claims/956830MCP
get_claim(id=956830)