whatmodel.sh / video / Cosmos-1.0-Autoregressive-4B
Cosmos-1.0-Autoregressive-4B
↗ share
+ contribute
Cosmos-1.0-Autoregressive-4B is an open-weights video model from NVIDIA, released January 2025. Its published size is 4B parameters. Licence: nvidia-open-model.
Catalogue identity notes
Comparison-only, proposed so the Cosmos paper's 3D-consistency table counts. The paper's 'Cosmos-Predict1-4B' is the 4B autoregressive base model (Table 10); weights nvidia/Cosmos-1.0-Autoregressive-4B (sha 0b168630ad2400affefe460f2743c071de1fb257, created 2025-01-07).
$ how it compares
The video models ranked either side of Cosmos-1.0-Autoregressive-4B, out of 139 with a score, then more from NVIDIA.
| # | model | score | where | ||
|---|---|---|---|---|---|
| 117 |  Kling 1.0 | 35±7 | comparison only | ||
| 118 |  VideoCrafter2 | 34±2 | local | ||
| 119 | Cosmos-1.0-Autoregressive-4B | 34±5 | comparison only | ||
| 120 |  Pika 1.0 | 32±3 | comparison only | ||
| 121 |  Open-Sora 1.1 | 32±7 | local | ||
| more from NVIDIA | |||||
| 38 | Cosmos3-Super | 58±4 | local | ||
| 60 | Cosmos3-Nano | 52±3 | local | ||
| 72 | Cosmos-1.0-Diffusion-7B-Video2World | 49±5 | comparison only | ||
$ what it's good at
text to video▇▅▄▆█▆▇▂▃▃▂▄ #75 of 94 * · 91% of its score · 2 tests
| test | result | vs the others in its run | counts |
|---|---|---|---|
| Cosmos 3D consistency: camera-pose estimation success rate on generated videos lab | 35.6% ✎ | █·█·······██ beats 1/3 | 87% |
| Cosmos 3D consistency: Sampson epipolar error on generated videos (lower is better) lab | 0.4330 ✎ | █···█···█··█ beats 1/3 | 4.2% |
$ where to run it
No verified API or download recorded yet · add one
Official weights and documentation: huggingface.co/nvidia ↗
$ how the score was worked out · Cosmos World Foundation Model… 91%
Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works
| source | results | share of the score |
|---|---|---|
 Cosmos World Foundation Model Platform for Physical AI (arXiv:2501.03575v3) · lab-reported | 2 | 91% |
| prior · pull toward the middle | 9% | |
| weighted average | = 34.1 |
$ improve this page
Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.
data updated · sources and licences · json