LongCat-Video: what the repository publishes
LongCat-Video publishes 13.6B parameters under the MIT License and states 720p output at 30fps. It states no GPU memory figure and names no card, so the one cell that decides whether the weights can be run on hardware somebody owns is empty. As of 2026-09-22.
| Field | What is stated | Detail |
|---|---|---|
| Licence | MIT License | Stated in the repository |
| Size | 13.6B parameters | Stated in the repository |
| Hardware | No memory figure recorded | The repository states none and names no card |
| Resolution | 720p | Stated with a 30fps frame rate |
| Length | No ceiling recorded | Described as minutes-long, with no figure attached |
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.
1Licence and size answered, cost unanswered
MIT and 13.6B are both one-line facts, quickly checked and quickly cleared. Neither tells a reader whether the model will load on the card in front of them, and the repository supplies nothing that does.
This is the most common shape of gap in this register, and it is worth stating as a pattern rather than as a complaint. A licence name and a parameter count are cheap to publish and interesting to announce. A memory figure requires running the thing and committing to a configuration, so it is the value most often left out.
2Minutes-long is a capability claim with no figure behind it
The repository describes the model as able to generate minutes-long videos and attaches no maximum to the phrase. The length cell therefore records that no ceiling is stated, rather than reading a number into an adjective.
The distinction matters because a ceiling is a commitment and an adjective is not. Three families here put an actual number on sixty seconds; a description of duration without a figure is a different kind of statement, and flattening the two into one column would make this register look more decided than it is.
3A 720p claim with nothing behind it in the next cell
720p at 30fps is the stated output, and nothing in the repository ties that output to a memory figure or to a card. So the resolution cell is filled and the hardware cell beside it is empty, which is the pairing a reader should treat most carefully.
An output size is the easiest value to publish and the least constraining, because it says what the code will emit rather than what emitting it costs. Read on its own it gets taken as evidence that the model is practical at that size, and this register keeps the two cells adjacent to make that harder to do.
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.
- LicenceMIT Licensestated in the repository
- Model size13.6B parametersstated in the repository
- Resolution720p, 30fps videosstated with a frame rate and no memory figure beside it
5Sources
Read from github.com/meituan-longcat 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: Cosmos Predict 2.5. Next family: HunyuanVideo.