H3 Max
#9 in Text-to-VideoFal · max · siet 27. August 2026 · 5× · tolest 01. Sept. 2026
67
Momentum
H3 Max is a post-trained video model by fal Research, built on the open-weight MiniMax H3 base model and released by fal on August 27, 2026. It is optimized for maximum inference speed, generating a 5-second 768p video clip with synchronized audio in under 3 seconds — roughly 35x faster than the official MiniMax H3 endpoint, according to fal. The model is available exclusively through fal as an API and playground (text-to-video and image-to-video); fal has not released standalone H3 Max weights. Unlike the base MiniMax H3 model (up to 2K resolution), H3 Max is capped at 768p.
Momentum-Verloop
03.06.01.09.
Features
| Fine-Tuning | No fine-tuning documented for H3 Max itself; LoRA training available for base model MiniMax H3 |
| Generation Time | 5-second clip at 768p rendered in under 3 seconds |
| License | No standalone weights released; API/playground access only, no model card or checkpoint available |
| Max Resolution | 768p (default/target resolution); 480p alternative; for 2K use standard MiniMax H3 |
| Max Video Length | 5 to 15 seconds |
| Platform | fal API, fal Playground/Agent (h3-max.com as interface) |
| Price | $0.06/sec at 768p (list price); launch promo $0.04/sec (until ~Sept 1, 2026); 15-sec clip approx. $0.90 |
| Release Date | August 27, 2026 |