MiniMax-H3 Physical-Reasoning Test: 42% Overall, Audio Just 27%
NUS-led team tested MiniMax-H3 on 517 multimodal tasks: 41.97% overall, video best at 56%, audio worst at 27.4%.
Since MiniMax open-sourced H3 in August 2026 the pace has been relentless: one unified omni-modal model, same-day ComfyUI support down to an RTX 3060, a 3-hour Day-0 port to Moore Threads GPUs, fal's post-trained H3 Max at 35x throughput, and FastH3 distillation cutting inference cost further. This topic tracks the full H3 chain — release, open weights, distillation, deployment.
NUS-led team tested MiniMax-H3 on 517 multimodal tasks: 41.97% overall, video best at 56%, audio worst at 27.4%.
Hybrid attention on MiniMax H3: local softmax plus frame-wise linear memory denoises a 14.4s 768p video in 6.9s on 8 B200s — faster than…
FastVideo open-sources FastH3 Preview v1: DMD2 distillation cuts MiniMax H3's 49 DiT calls to 4, plus 90% sparse attention — up to 14x on…
fal's H3 Max: post-trained MiniMax H3 makes 5s clips in ~3s, ~35x official throughput, #1 in preference and Design Arena evals; capped at…
After MiniMax H3's weights landed on Hugging Face on August 3, community testing revealed that the open-source release only ships the base…
On August 3, Meitu's RoboNeo announced its integration with MiniMax's newly open-sourced multimodal generation model MiniMax H3, focusing…
MiniMax has introduced H3, a general-purpose omni-modal video model that accepts text, images, video and audio in one context and generates…
On 3 August 2026 MiniMax released the open-weights MiniMax H3 (Hailuo 3.0) omni-modal video model, and ComfyUI shipped Day-0 support that…
On the day MiniMax open-sourced H3, Moore Threads said it achieved stable deployment on a single eight-card MTT S5000 node in only three…