Primary use stated or inferred cautiously from official documentation.
N
NVIDIA / nvidia/Cosmos3-Nano-Sim-Bimanual
Cosmos3-Nano-Sim-Bimanual
NVIDIA causal model for visually simulating bimanual manipulation and camera navigation, published on October 20, 2026. It receives an initial image, robot actions, camera trajectory, and optional text. Its output is video, not actions: it should not be confused with a control policy. The checkpoint declares an AgiBot body; other morphologies require specific validation.
01 / SUMMARY
The essential
Maximum input capacity when published by the source.
Documented maximum generation limit.
Verified availability channels.
Primary functional family and verified specialties.
Declared terms for API access, model weights, or self-hosting.
02 / TECHNICAL SHEET
Limits and integration
| API ID | nvidia/Cosmos3-Nano-Sim-Bimanual |
|---|---|
| Model type | Robotics · Video · Multimodal |
| Access model | Open weights |
| License | OpenMDW 1.1 |
| Deployment | Self-hosted |
| Release | 20 OCT 2026 |
| Knowledge cutoff | Not published |
| Entrance | Text · Image · Actions |
| Exit | Video |
| Context window | Initial image, actions, and camera trajectory |
| maximum output | Video at 480p; playback at 30 fps |
| Reasoning | Not published |
| Published tools | |
| Structured Outings | Not published |
| Batch processing | Not published |
| Prompt cache | Not published |
| Fine-tuning | Not published |
| Verified platforms | Hugging Face · PyTorch |
Exploration of visual consequences of actions and evaluation of closed-loop navigation. History and causal fragments allow interleaving controls and generation. The declared resolution is 480p with playback at 30 fps. We do not inherit the 600-frame maximum from Sim-Transfer: we have not verified an equivalent universal maximum for this variant.
The ability to represent movements does not guarantee that it respects collisions, forces, or robot limits. Preserve camera calibration and the definition of each action dimension. Review object changes, drift, and cumulative errors. Do not execute actions on real hardware solely because the video seems plausible; supervision and safety limits are independent.
03 / CAPABILITIES
What it can do
Orientation
Exploration of visual consequences of actions and evaluation of closed-loop navigation. History and causal fragments allow interleaving controls and generation. The declared resolution is 480p with playback at 30 fps. We do not inherit the 600-frame maximum from Sim-Transfer: we have not verified an equivalent universal maximum for this variant.
Context
Initial image, actions, and camera trajectory · Video at 480p; playback at 30 fps
Tools and integration
Not published
Access
Hugging Face · downloadable checkpoint
Input modalities
TextImageActions
04 / PRICES
Documented cost
| Concept | Worth | Unit/condition |
|---|---|---|
| Standard input | No published official API rate | Standard API rate |
| Cached input | Not published | Reading reused prefixes |
| Cache write or storage | Not published | The condition varies by provider |
| Standard output | Self-hosted infrastructure cost | May include reasoning tokens |
| Batch input | Not published | Asynchronous processing |
| Batch output | Not published | Asynchronous processing |
Public 16B parameter checkpoint for self-hosted infrastructure. We have not verified a commercial endpoint or token rates: numerical costs are left blank. Downloadable weights do not make inference free. Budget for GPU, memory, video encoding, and operation; review the license of each variant and its dependencies. Playback frames per second do not equate to generation performance.
i Prices change and may depend on level, region or context length. Check the source before making a decision.
05 / EVALUATIONS
How to read the results
“
The model card separates PAI-Bench-G I2V, Sana-WM navigation, and AgiBot manipulation. For PAI-Bench-G, we publish video quality and ATE after the described alignment protocol. A navigation score does not certify manipulation nor should it be averaged with PSNR. The results come from NVIDIA and have not been reproduced by Inferama.
| Benchmark | Result | Metric | Source |
|---|---|---|---|
| PAI-Bench-G I2V Video Quality ↑ | 73,9099 | Score | View source ↗ |
| PAI-Bench-G I2V ATE ↓ | 4,7072 | Score | View source ↗ |
Inferama only highlights a “best result” when the metric, test set, configuration, and date allow for an equivalent comparison. The supplier's figures are presented as claims from its own source.
06 / VERSIONS
Chronology
Cosmos3-Nano-Sim-Bimanual
Version added to Inferama's verified catalog.
07 / SOURCES