Pyramid Flow: what the repository publishes
Pyramid Flow states its output precisely and its cost not at all: 10-second videos at 768p and 24 FPS, with a 384p checkpoint producing 5 seconds at the same rate. The card carries the stabilityai-ai-community tag, no parameter count and no memory figure. As of 2026-09-22.
| Field | What is stated | Detail |
|---|---|---|
| Licence | stabilityai-ai-community | A licence name from one company on another's release |
| Size | No parameter count recorded | The card states none |
| Hardware | No memory figure recorded | Offloading is discussed without gigabytes |
| Resolution | 768p, with a 384p checkpoint | Stated per checkpoint |
| Length | 10 seconds at 768p, 5 at 384p | Seconds, at 24 FPS, for both checkpoints |
Inclusion rule. The same five fields for every family here. A field the repository does not state is recorded as not stated rather than estimated from the parameter count. Order. Fixed field order, identical on every family page.
1Two cells full, two cells empty, and the empty ones are the expensive ones
The length and resolution values here are among the most complete in this register: a duration, a resolution and a frame rate for each of two published checkpoints. The size and hardware values hold nothing, because the card holds nothing.
That distribution deserves naming because it is exactly backwards for provisioning. A reader can tell precisely what comes out and has no published route to working out what getting it costs, so the only honest next step is to try the thing rather than to plan around it.
2A community licence written by one company, applied to another's release
The tag is stabilityai-ai-community: a licence drafted by one company, appearing on a release published by somebody else. This register records the name on the card and nothing past it, because who wrote a licence and who applied it are two separate facts and only the second one is visible from here.
The practical effect matches any other vendor-written licence in this register. The text has to be read rather than recognised, and a reader who cleared Apache 2.0 or MIT on another entry here has cleared nothing that applies to this one.
3Offloading mentioned with no figure to attach it to
The card discusses CPU offloading as a way to save memory, and mentions that with enough GPU memory a decoding setting can be switched the other way for speed. Neither sentence carries a number, so neither becomes a value in the hardware cell.
One other entry here states the same trade-off in gigabytes on both sides of the switch. The difference between the two cards is not what the models do; it is what the publishers wrote down, and what was written down is the only thing a register can record.
4The same field, across every family
| Family | Licence |
|---|---|
| Allegro | apache-2.0 |
| CogVideoX | Split by model size |
| Cosmos Predict 2.5 | NVIDIA Open Model License |
| FramePack | Apache 2.0 |
| HunyuanVideo | tencent-hunyuan-community |
| LongCat-Video | MIT License |
| LTX-Video | Apache 2.0 |
| MAGI-1 | Apache License 2.0 |
| Mochi 1 | Apache 2.0 |
| Open-Sora | Apache 2.0 |
| Pyramid Flow | stabilityai-ai-community |
| SkyReels-V2 | skywork-license |
| Stable Video Diffusion | stable-video-diffusion-community |
| Step-Video-T2V | MIT |
| Wan | Apache 2.0 |
Inclusion rule. Families whose vendor has published weights openly. A family whose repository does not state this particular field keeps its row, marked hollow. Order. Alphabetical by family name.
| Family | Size |
|---|---|
| Allegro | VAE 175M and DiT 2.8B |
| CogVideoX | 2B and 5B, carried in the model names |
| Cosmos Predict 2.5 | 2,059,174,912 parameters |
| FramePack | 13B models |
| HunyuanVideo | Over 13 billion parameters |
| LongCat-Video | 13.6B parameters |
| LTX-Video | A 13B model |
| MAGI-1 | 24B and 4.5B models |
| Mochi 1 | 10 billion parameters |
| Open-Sora | An 11B model |
| Pyramid Flow | No parameter count recorded |
| SkyReels-V2 | 1.3B, 5B and 14B variants |
| Stable Video Diffusion | 2B params |
| Step-Video-T2V | 30 billion parameters |
| Wan | A 5B text-to-video and image-to-video model |
Inclusion rule. Families whose vendor has published weights openly. A family whose repository does not state this particular field keeps its row, marked hollow. Order. Alphabetical by family name.
| Family | Hardware |
|---|---|
| Allegro | 9.3G BF16 with cpu_offload |
| CogVideoX | From 4GB in FP16 on the 2B |
| Cosmos Predict 2.5 | 32.54 GB of GPU VRAM |
| FramePack | 6GB minimum for 1800 frames at 30fps |
| HunyuanVideo | 60GB peak at 720px1280px, 129 frames |
| LongCat-Video | No memory figure recorded |
| LTX-Video | No minimum stated |
| MAGI-1 | H100 or H800 x 8 for the 24B |
| Mochi 1 | About 60GB VRAM on a single GPU |
| Open-Sora | 60.3GB peak at 768px on one GPU |
| Pyramid Flow | No memory figure recorded |
| SkyReels-V2 | About 14.7GB peak on the 1.3B at 540P |
| Stable Video Diffusion | About 180s on an A100 80GB card |
| Step-Video-T2V | 78.55 GB peak at 768px768px204f |
| Wan | A consumer 4090 on the 5B release |
Inclusion rule. Families whose vendor has published weights openly. A family whose repository does not state this particular field keeps its row, marked hollow. Order. Alphabetical by family name.
- Licencestabilityai-ai-communitythe tag on the model card
- Length10-second videos at 768p resolution and 24 FPS, with a 384p checkpoint that generates 5-second video at 24FPSstated per checkpoint with the frame rate
- Resolution768p on one checkpoint and 384p on the otherstated per checkpoint
5Sources
Read from huggingface.co/rain1011 on 2026-09-22. Hardware figures across every family are collected on requirements; the licence column is set out on licence and what counts as a published weight is on how read. Previous family: FramePack. Next family: SkyReels-V2.