HappyHorse-1.0

Alibaba ATH, open-sourced 2026-04 under Apache 2.0 — commercially usable. This is the weights release, not the metered HappyHorse 1.1 API above. At 15B parameters, VRAM is the hard gate: full FP16 needs 30GB+, so consumer cards (12 / 16 / 24GB) must use INT8 / GGUF quantisation or CPU offload, at a speed cost.

32 GB

VRAM floor

15B

Parameters

open-weight models

Open source Highest bar

Weights are free, hardware is not: you buy the GPU up front (one-time estimate, may rise) and still pay for power and time. Six of the twelve are Apache 2.0 (HappyHorse-1.0, SANA-WM, Bernini, MAGI-2, LingBot-Video, LongLive 2.0); LTX-2 and LTX-2.5 use community licences and SANA-Video uses the NVIDIA Open Model License. Hardware runs from one RTX 5090 to 8× H100. See all 12 →

Hardware requirements, as published

VRAM floor: 30GB+ in full FP16
Official hardware: Single card with 30GB+; 12 / 16 / 24GB cards need INT8 / GGUF quantisation or CPU offload
Output: 1080P / 15 sec with joint audio-video generation
Size: 15B parameters
Licence: Apache 2.0 (commercially usable)

Full FP16 needs 30GB+ of VRAM; consumer cards (12 / 16 / 24GB) must use INT8 / GGUF quantisation or CPU offload, at a speed cost.Alibaba ATH · happyhorse.com

1 records · showing Specs checked 2026-11-10 · hardware 2026-10-21

#	Platform / model	Layer	VRAM · setup	One-time hardware
—	OSSHappyHorse-1.0 ↗Alibaba ATH · 2026-04 · Apache 2.0 · 15B	Open source	32 GB30GB+ in FP16 · 24GB needs quant	≈ ¥35,000 · one-time1× RTX 5090 32GB · estimate, may rise

Other models in this layer

LTX-2.5 SANA-WM Bernini MiniMax H3-Base MAGI-2 Preview LTX-2 SANA-Video LongLive 2.0 Matrix-Game 4.0 SkyReels-V3 LingBot-Video

⚖

Data on this page is indicative only — listing a vendor is not a recommendation or guarantee.Figures are compiled and converted from public pages and can change at any time; check the vendor's current pricing before you buy. Read the disclaimer