Integrate reviewed Cosmos3-Edge-FP8 model card with current usage and performance

Use the reviewed MR overview, quality benchmarks, and training/evaluation datasets; retain current HF usage examples and performance measurements. Add the PNG relationship diagram, set release date to November 9, 2026, and correct precision and section links. Exclude internal-only sections. Existing shared subcards are unchanged.

README.md

@@ -17,47 +17,25 @@ tags:
 
 # Model Overview: Cosmos3-Edge-FP8
 
+![Cosmos3-Edge base model and its FP8-quantized variant](images/cosmos3_edge_fp8.png)
+
+The highlighted checkpoint, Cosmos3-Edge-FP8, is an FP8-quantized version of the Cosmos3-Edge base model.
+
 ## Description
 
+Cosmos3-Edge-FP8 is an FP8-quantized variant of Cosmos3-Edge. It retains the base model architecture and intended applications, with FP8 quantization applied to selected model components.
+
 Cosmos3 is a collection of Omnimodal world models capable of generating dynamic, high-quality video, image, audio, and action commands from combinations of text, image, video, and action trajectory inputs. It serves as a foundational building block for a broad range of Physical AI applications and research spanning world understanding, world generation, simulation, and embodied policy learning.
 
 This model is ready for commercial and non-commercial use.
 
-> **Update — August 25, 2026:** The Cosmos3-Edge-FP8 generator checkpoint, runtime defaults, usage examples, and benchmark results have been updated. Users pulling from `main` should refresh their local snapshot. See the [update announcement](https://huggingface.co/nvidia/Cosmos3-Edge-FP8/discussions/62) for compatibility and reproducibility details.
-
 **Model Developer:** NVIDIA
 
 ### Model Versions
 
-**Released on: 07/20/2026**
 - Cosmos3-Edge-FP8:
   - Given multimodal inputs including text, images, video, and action trajectories, generate coherent text, images, video, and action outputs for multimodal understanding, world simulation, future prediction, action reasoning, and Physical AI applications.
 
-- Cosmos3-Edge-Policy-DROID:
-  - Given language instructions and visual observations from the DROID robot platform, generate robot action trajectories for manipulation and control tasks.
-
-- Cosmos3-Super-Image2Video-4Step:
-  - Given one or more input images and optional text instructions, generate temporally coherent video sequences that are consistent with the provided visual content. Distilled from Cosmos3-Super-Image2Video using [Improved Distribution Matching Distillation (DMD2)](https://arxiv.org/abs/2405.14867), enabling high-quality generation in 4 steps.
-
-- Cosmos3-Super-Text2Image-4Step:
-  - Given text input, generate high-fidelity images that are consistent with the provided description. Distilled from Cosmos3-Super-Text2Image using [Improved Distribution Matching Distillation (DMD2)](https://arxiv.org/abs/2405.14867), enabling high-quality generation in 4 steps.
-
-**Released on: 05/31/2026**
-- Cosmos3-Nano:
-  - Given multimodal inputs including text, images, video, audio, and action trajectories, generate coherent text, images, video, audio, and action outputs for multimodal understanding, world simulation, future prediction, action reasoning, and Physical AI applications.
-
-- Cosmos3-Super:
-  - Given multimodal inputs including text, images, video, audio, and action trajectories, generate coherent text, images, video, audio, and action outputs for multimodal understanding, world simulation, future prediction, action reasoning, and Physical AI applications.
-
-- Cosmos3-Nano-Policy-DROID:
-  - Given language instructions and visual observations from the DROID robot platform, generate robot action trajectories for manipulation and control tasks.
-
-- Cosmos3-Super-Image2Video:
-  - Given one or more input images and optional text instructions, generate temporally coherent video sequences that are consistent with the provided visual content.
-
-- Cosmos3-Super-Text2Image:
-  - Given text input, generate high-fidelity images that are consistent with the provided description.
-
 ### License
 
 This model is released under the [OpenMDW1.1](https://openmdw.ai/license/1-1/)
@@ -72,8 +50,8 @@ Physical AI: Encompassing robotics, autonomous vehicles (AV), and smart space en
 
 ### Release Date
 
-Hugging Face 07/20/2026 via [https://huggingface.co/collections/nvidia/cosmos3](https://huggingface.co/collections/nvidia/cosmos3)
-GitHub 07/20/2026 via [https://github.com/nvidia/cosmos](https://github.com/nvidia/cosmos)
+Hugging Face 11/09/2026 via [https://huggingface.co/collections/nvidia/cosmos3](https://huggingface.co/collections/nvidia/cosmos3)
+GitHub 11/09/2026 via [https://github.com/nvidia/cosmos](https://github.com/nvidia/cosmos)
 
 ## Model Architecture
 
@@ -87,18 +65,8 @@ Cosmos3 is an Omni-modal foundation model built on a Mixture-of-Transformers (Mo
 
 **Number of trainable model parameters:**
 
-**Released on: 07/20/2026**
 - Cosmos3-Edge-FP8: 4B
-- Cosmos3-Edge-Policy-DROID: 4B
-- Cosmos3-Super-Image2Video-4Step: 64B
-- Cosmos3-Super-Text2Image-4Step: 64B
 
-**Released on: 05/31/2026**
-- Cosmos3-Nano: 16B
-- Cosmos3-Super: 64B
-- Cosmos3-Nano-Policy-DROID: 16B
-- Cosmos3-Super-Image2Video: 64B
-- Cosmos3-Super-Text2Image: 64B
 
 ## Input/Output Specifications
 
@@ -188,7 +156,7 @@ Our AI models are designed and/or optimized to run on NVIDIA GPU-accelerated sys
 
 - Linux (We have not tested on other operating systems.)
 
-**Note:** Only BF16 precision is tested. Other precisions like FP4, FP8, and FP16 are not officially supported.
+**Precision:** This repository provides the FP8-quantized Cosmos3-Edge checkpoint. Components excluded from quantization retain their original precision. See [Performance Benchmark](#performance-benchmark) for backend- and hardware-specific precision details.
 
 The integration of foundation and fine-tuned models into AI systems requires additional testing using use-case-specific data to ensure safe and effective deployment. Following the V-model methodology, iterative testing and validation at both unit and system levels are essential to mitigate risks, meet technical and functional requirements, and ensure compliance with safety and ethical standards before deployment.
 
@@ -203,11 +171,11 @@ The integration of foundation and fine-tuned models into AI systems requires add
 - **Time period for testing data collection:** N/A (standard public benchmarks)
 - **Time period for validation data collection:** N/A (standard public benchmarks)
 
-Raw data from internal and external sources is transformed into training-ready data through multiple stages of curation, filtering, and quality review. Data acquisition spans diverse multimodal sources — robotics, autonomous driving, industrial environments, indoor and outdoor scenes, varied lighting and weather conditions, camera viewpoints, object categories, and human activities — to broaden coverage across Physical AI operating environments. Automated filtering pipelines remove corrupted, duplicate, low-quality, and restricted content. Metadata analysis, heuristic rules, and model-assisted classifiers are applied during preprocessing to flag anomalous distributions and low-diversity subsets. Human review supplements automated filtering for selected datasets, benchmark construction, and targeted quality analysis. Datasets are balanced across modalities and task categories — visual reasoning, text-to-image, text-to-video, image-to-video, video transfer, action-conditioned generation, and action command generation — to reduce overrepresentation of narrow domains. Synthetic and simulation-based augmentation supplements coverage of rare physical interactions and edge-case scenarios. Deduplication and provenance tracking are applied across the corpus. The resulting processed data is converted into model-ready tokenized or encoded representations through modality-specific preprocessors before training begins.
+Raw data from internal and external sources is transformed into training-ready data through multiple stages of curation, filtering, and quality review. Data acquisition spans diverse multimodal sources — robotics, autonomous driving, industrial environments, indoor and outdoor scenes, varied lighting and weather conditions, camera viewpoints, object categories, and human activities — to broaden coverage across Physical AI operating environments. Automated filtering pipelines remove corrupted, duplicate, low-quality, and restricted content. Metadata analysis, heuristic rules, and model-assisted classifiers are applied during preprocessing to flag anomalous distributions and low-diversity subsets. Human review supplements automated filtering for selected datasets, benchmark construction, and targeted quality analysis. Datasets are balanced across modalities and task categories — visual reasoning, text-to-image, text-to-video, image-to-video, audio generation, video transfer, action-conditioned generation, and action command generation — to reduce overrepresentation of narrow domains. Synthetic and simulation-based augmentation supplements coverage of rare physical interactions and edge-case scenarios. Deduplication and provenance tracking are applied across the corpus. The resulting processed data is converted into model-ready tokenized or encoded representations through modality-specific preprocessors before training begins.
 
-Training datasets passed through multiple layers of automated and manual safeguards designed to reduce the presence of harmful or policy-violating content across categories including weapons and weapons-related instructional content, criminal planning, child sexual abuse material (CSAM), non-consensual intimate imagery (NCII), sexual content involving minors, harassment, hate speech, profanity, threats and incitement to violence, self-harm or suicide-related content, and graphic violence. Data sources are reviewed for licensing compatibility, provenance, and alignment with internal data governance and safety policies before admission into training corpora. Automated filtering pipelines combine multiple detection strategies: hash-matching against known CSAM and NCII reference databases; classifier-based moderation models trained for explicit sexual content, hate speech, violence, weapons imagery, and other restricted categories; keyword and regex-based screening for criminal-planning, threats, and self-harm phrases in text data; metadata and provenance heuristics for source-level risk signals; and embedding-based anomaly detection to surface samples that fall outside expected distributions. Human review and targeted audits supplement automated filtering for selected datasets, benchmark construction, and safety-sensitive evaluation. For multimodal Physical AI data (robotics, autonomous driving, industrial scenes), additional filtering targets invalid action trajectories, physically implausible interactions, and unsafe control sequences. Synthetic and simulation-generated data are evaluated through internal validation before inclusion. Benchmark evaluations and red-team testing are applied post-training to surface remaining safety gaps across world generation, reasoning, and action tasks. No large-scale data-filtering process can guarantee complete removal of all harmful content; residual risks may remain, particularly in rare edge cases or open-world deployment settings. Ongoing monitoring and dataset review continue post-release.
+Training datasets passed through multiple layers of automated and manual safeguards designed to reduce the presence of harmful or policy-violating content across categories including weapons and weapons-related instructional content, criminal planning, child sexual abuse material (CSAM), non-consensual intimate imagery (NCII), sexual content involving minors, harassment, hate speech, profanity, threats and incitement to violence, self-harm or suicide-related content, and graphic violence. Data sources are reviewed for licensing compatibility, provenance, and alignment with internal data governance and safety policies before admission into training corpora. Automated filtering pipelines combine multiple detection strategies: hash-matching against known CSAM and NCII reference databases; classifier-based moderation models trained for explicit sexual content, hate speech, violence, weapons imagery, and other restricted categories; keyword and regex-based screening for criminal-planning, threats, and self-harm phrases in text data; metadata and provenance heuristics for source-level risk signals; and embedding-based anomaly detection to surface samples that fall outside expected distributions. Human review and targeted audits supplement automated filtering for selected datasets, benchmark construction, and safety-sensitive evaluation. For multimodal Physical AI data (robotics, autonomous driving, industrial scenes), additional filtering targets invalid action trajectories, physically implausible interactions, and unsafe control sequences. Synthetic and simulation-generated data are evaluated through internal validation before inclusion. Benchmark evaluations and red-team testing are applied post-training to surface remaining safety gaps across world generation, reasoning, audio, and action tasks. No large-scale data-filtering process can guarantee complete removal of all harmful content; residual risks may remain, particularly in rare edge cases or open-world deployment settings. Ongoing monitoring and dataset review continue post-release.
 
-- For more information about the datasets used to train this model, please see the [Public Summary of Training Content](https://docs.nvidia.com/cosmos/latest/_downloads/e482b7114ce8dbfbb07d2d4b42cafe4e/training-content-Cosmos-3.pdf).
+For more information about the datasets used to train this model, please see the [Public Summary of Training Content](https://docs.nvidia.com/cosmos/latest/cosmos3/index.html).
 
 **Data Modality and Training Data Size**
 
@@ -216,6 +184,7 @@ Training datasets passed through multiple layers of automated and manual safegua
 | Text     | 22M                 | Not Applicable       |
 | Image    | 19M                 | 767M                 |
 | Video    | 1M                  | 348M                 |
+| Audio    | Not Applicable      | 139M                 |
 | Action   | Not Applicable      | 7M                   |
 
 **Data Collection Method by dataset**
@@ -224,9 +193,9 @@ Training datasets passed through multiple layers of automated and manual safegua
 
 **Labeling Method by dataset**
 
-- Hybrid: Human, Automated
+- Hybrid: Manually-Labelled, Automated
 
-**Properties:** The training, testing, and evaluation datasets consist of diverse multimodal video, image, action, synthetic, and sensor-conditioned data sourced from NVIDIA-owned data and publicly available, commercially permissive datasets. These datasets are curated to exclude known restricted content and to support building an Omni model that learns to generate and reason about dynamic physical environments across world reasoning and generation tasks.
+**Properties:** The training, testing, and evaluation datasets consist of diverse multimodal video, image, audio, action, synthetic, and sensor-conditioned data sourced from NVIDIA-owned data and publicly available, commercially permissive datasets. These datasets are curated to exclude known restricted content and to support building an Omni model that learns to generate and reason about dynamic physical environments across world reasoning and generation tasks.
 
 ### Public Datasets
 
@@ -262,39 +231,42 @@ Training datasets passed through multiple layers of automated and manual safegua
 
 **Labeling Method by dataset**
 
-- Hybrid: Human, Automated
-
-**Properties:** The training, testing, and evaluation datasets consist of diverse multimodal video, image, action, synthetic, and sensor-conditioned data sourced from NVIDIA-owned data and publicly available, commercially permissive datasets. These datasets are curated to exclude known restricted content and to support building an Omni model that learns to generate and reason about dynamic physical environments across world reasoning and generation tasks.
-
-## Benchmarks
-
-For detailed evaluations of the base model, see our [technical paper](https://research.nvidia.com/labs/cosmos-lab/cosmos3/technical-report.pdf).
-
-### Overall
-
-The table below summarizes Cosmos3-Edge-FP8 across reasoning and generation. Each reasoning column (General, Robotics, Smart Infrastructure, Driving) reports the average score over that capability's benchmarks. For generation, **Image2Video** is the PAIBench overall score and **Policy: Robot** is the RoboLab success rate. In each column, the best result is in **bold** and the second-best is <u>underlined</u>.
-\* denotes post-trained Cosmos3 variants: [Cosmos3-Nano-Policy-DROID](https://huggingface.co/nvidia/Cosmos3-Nano-Policy-DROID) and [Cosmos3-Edge-Policy-DROID](https://huggingface.co/nvidia/Cosmos3-Edge-Policy-DROID#robolab).
+- Hybrid: Manually-Labelled, Automated
 
-![Overall benchmark results](https://huggingface.co/nvidia/Cosmos3-Edge/resolve/main/assets/benchmark-overall.png)
+**Properties:** The training, testing, and evaluation datasets consist of diverse multimodal video, image, audio, action, synthetic, and sensor-conditioned data sourced from NVIDIA-owned data and publicly available, commercially permissive datasets. These datasets are curated to exclude known restricted content and to support building an Omni model that learns to generate and reason about dynamic physical environments across world reasoning and generation tasks.
 
-### Reasoning Benchmarks
+## Quality Benchmark
 
-![Reasoning benchmarks](https://huggingface.co/nvidia/Cosmos3-Edge/resolve/main/images/benchmark-reasoning.png)
+### Reasoner Benchmarks
 
-### Generation Benchmarks
+We compare **Cosmos3-Edge-FP8** with its BF16 counterpart across general reasoning, robotics, smart infrastructure, and driving benchmarks. Each domain score is the unweighted mean of the included benchmark scores. The benchmark list below defines the evaluation scope. Benchmarks without FP8 results are excluded.
 
-#### Image-to-Video Generation
+| Domain | BF16 | FP8 |
+| --- | --- | --- |
+| General ↑ | 56.56 | 58.10 |
+| Robotics ↑ | 49.17 | 48.68 |
+| Smart Infrastructure ↑ | 52.00 | 51.95 |
+| Driving ↑ | 62.50 | 62.38 |
 
-All models are evaluated on image-to-video generation at **480p, 24 fps**. Throughput is the number of frames generated per second, measured in eager mode on a single **NVIDIA H100** GPU. Cosmos3-Edge-FP8 delivers the highest generation throughput while achieving competitive quality across PAIBench, RBench, and PhysicsIQ.
+- **General** (18 benchmarks): MMBench_DEV_EN_V11, RealWorldQA, CVBench, VideoPhy2, CausalVQA_4fps, MVPBench, CountBenchQA, AI2D_TEST, DocVQA_VAL, InfoVQA_VAL, OCRBench_v2, LogicVista, MMMU_Pro_10c, MVBench_2fps64f, BlinkSpatial, BlinkDepth, HallusionBench, IFBench
+- **Robotics** (10 benchmarks): CosmosReason, RefSpatial, VSI-Bench_32frame, SparBench, MMSIBench_wo_circular, MMSIVideoBench_32frame, HealthSurgiBench, CosmosERQA, RoboSpatialHome, Where2Place
+- **Smart Infrastructure** (8 benchmarks): VANTAGE_2DGrounding, VANTAGE_Astro2D, VANTAGE_2DPointing, VANTAGE_DVC, VANTAGE_EventVerification, VANTAGE_SOT, VANTAGE_Temporal, VANTAGE_VQA
+- **Driving** (3 benchmarks): LingoQA, AVSpecialCollisionBench, AVSpecialStopBehaviorBench
 
-![Image-to-Video benchmark results](https://huggingface.co/nvidia/Cosmos3-Edge/resolve/main/assets/benchmark-image2video.png)
+### Generator Benchmarks
 
+We compare **Cosmos3-Edge-FP8** with its BF16 counterpart. PAIBench-G evaluates text-to-video and image-to-video generation. RBench assesses task correctness, physical plausibility, and visual fidelity in robotics videos. PhysicsIQ compares video continuations with recorded physical outcomes.
 
-#### Action
+| Benchmark | BF16 | FP8 |
+| --- | --- | --- |
+| PAIBench-G I2V ↑ | 81.10 | 80.75 |
+| PAIBench-G T2V ↑ | 76.90 | 76.58 |
+| RBench ↑ | 55.10 | 56.52 |
+| PhysicsIQ ↑ | 31.28 | 31.37 |
 
-The Edge model is a strong initialization for downstream action tasks. For example, post-training it on the DROID dataset produces a policy whose RoboLab success rate is reported in the [Cosmos3-Edge-Policy-DROID model card](https://huggingface.co/nvidia/Cosmos3-Edge-Policy-DROID#robolab).
+*For quality benchmarks, BF16 results were obtained using PyTorch. FP8 results were obtained using vLLM-Omni.*
 
-### Performance Benchmarks
+## Performance Benchmark
 
 The table reports single-GPU inference performance for Cosmos3-Edge-FP8 using FLASH_ATTN unless otherwise noted.
 
@@ -991,9 +963,9 @@ Cosmos3 outputs should not be treated as physically accurate simulation, reliabl
 
 ## Inference
 
-**Acceleration Engine:** [vLLM-Omni](https://github.com/vllm-project/vllm-omni), [vLLM](https://github.com/vllm-project/vllm), [PyTorch](https://pytorch.org/)
+**Acceleration Engine:** [vLLM-Omni](https://github.com/vllm-project/vllm-omni), [vLLM](https://github.com/vllm-project/vllm), [PyTorch](https://pytorch.org/), [Hugging Face Diffusers](https://github.com/huggingface/diffusers)
 
-**Test Hardware:** H100, H20, RTX PRO 6000, DGX Station, DGX Spark, Jetson Thor, Jetson AGX Orin
+**Test Hardware:** See [Performance Benchmark](#performance-benchmark) for the evaluated GPU configurations, including emulated T3000 results and Orin weight-only W8A16 compute.
 
 ## Ethical Considerations