Models & Labs
NVIDIA Cosmos 3 Open Models for World Simulation
Hugging Face Bloghigh confidence
Why it matters
- →Provides open-weight world models specifically tuned for synchronized LiDAR and camera generation in autonomous driving.
- →Enables closed-loop robotics policy training through Sim-Bimanual, which predicts visual consequences of robot actions.
- →Offers a standardized foundation for researchers to adapt simulation environments without building generative models from scratch.
NVIDIA has released four new open-weight checkpoints for its Cosmos 3 Nano model family on Hugging Face, designed for interactive world simulation in autonomous driving and robotics. Each model contains 16B trainable parameters and is licensed under OpenMDW1.1 for commercial and non-commercial use. The suite includes Transfer-Auto for generating multi-view driving data with varied weather, Sim-Auto for streaming synchronized camera and LiDAR observations, Sim-Transfer for rendering video from physics simulator controls, and Sim-Bimanual for predicting video from robot actions. These models utilize a Mixture-of-Transformers backbone and require NVIDIA GPUs for inference.
Read originalThe story around this
Earlier coverage that leads up to this article, and what followed. Lines connect each piece to the closest one after it, converging here.
How we got here
- 1NVIDIA Releases Cosmos 3 Edge for Robotics
Hugging Face Blog · July 20, 2026 · Same story
- 2NVIDIA Unveils Real-Time Surgical Simulator
Hugging Face Blog · July 27, 2026 · Related
- 3NVIDIA Cosmos 3 Advances Open World Models for Physical AI
NVIDIA Blog · August 6, 2026 · Same story
- 4NVIDIA Boosts Local AI with New Open Models
NVIDIA Blog · August 11, 2026 · Same story
- 5Nvidia Highlights Importance of AI Harness Over Model
TechCrunch AI · August 21, 2026 · Background
- 6Worldmodeldata brokers video game data for AI training
WIRED AI · September 28, 2026 · Related
- 7Researchers Drive Car with GPT-6 Astra
WIRED AI · October 7, 2026 · Background
- 8Nvidia releases open Nemotron models for agents and PCs
TechCrunch AI · October 19, 2026 · Related
More from Hugging Face Blog
© Hugging Face BlogResearchagents
ToyBlocks: Controlled simulation for robot learning research
Hugging Face launched ToyBlocks, a controlled simulation framework designed to isolate and study the specific mechanics of robot learning. By focusing on a simple tabletop environment with blocks, it allows researchers to automatically generate massive datasets of tasks and demonstrations without manual teleoperation. This setup enables precise testing of how high-level reasoning models like Molmo 2 interact with low-level action controllers like ManiFlow. The real value lies in reproducibility; scientists can now systematically vary training data to understand exactly why a robot fails or succeeds, moving beyond the black-box nature of current robotics benchmarks.
Hugging Face Blog
© Hugging Face BlogResearchother
Open ASR Leaderboard adds Egyptian and Nigerian English
The Open ASR Leaderboard is expanding its evaluation scope by adding private test sets for Egyptian and Nigerian English, addressing a significant gap in global representation. This update moves beyond the traditional Inner Circle accents to include Outer and Expanding Circle varieties, providing a more realistic benchmark for speech recognition models. By partnering with Liva AI for data collection, Hugging Face ensures these evaluations reflect natural, high-context user interactions rather than artificial tasks. This shift forces developers to optimize for diversity, making model performance metrics more trustworthy and globally relevant.
Hugging Face Blog
© Hugging Face BlogModels & Labsmusic
Phonon TTS: High-Quality On-Device Speech
Hugging Face and Gradium have released Phonon, a text-to-speech model that runs entirely on-device without network connectivity. At just 31 million parameters, it tops the Open TTS Leaderboard for speed and ranks in the top three for intelligibility across five languages, outperforming models up to 100 times its size. The release includes Apache 2.0 licensed weights for commercial use and a new inference framework supporting CPU, GPU, and NPU backends. This shifts high-fidelity voice synthesis from cloud dependency to local execution, enabling privacy-preserving applications on phones and browsers.
Hugging Face Blog
More in Models & Labs
Models & Labsother
llama.cpp b12177 adds CUDA 13 and ROCm 10
This release quietly closes the hardware gap for local inference by adding default support for CUDA 13.5 and ROCm 10.0 alongside existing CUDA 12 builds. AMD GPU users finally get first-class parity with NVIDIA, while Snapdragon mobile chips gain dedicated NPU acceleration paths. Apple Silicon KleidiAI is explicitly disabled here, a notable regression for Mac performance that suggests ongoing tuning issues. The inclusion of OpenVINO and SYCL backends further cements llama.cpp as the universal runtime for inference across nearly every major silicon architecture.
llama.cpp Releases
Models & Labsmodels
llama.cpp adds Kimi K3.1-Pro support
Kimi’s new K3.1-Pro model joins the llama.cpp ecosystem, bringing a distinct architectural twist to local inference. By replacing full-attention MLA layers with DSA (Dynamic Sparse Attention) and linear-attention KDA layers, this release tests how well sparse caching handles hybrid attention patterns. The implementation keeps indexer precision high to avoid quality loss during top-k selection, ensuring the model runs accurately on consumer hardware. This expands the library’s reach into newer Chinese open-weight models that rely on complex memory management.
llama.cpp Releases
Models & Labsother
llama.cpp b12182 release with CUDA 13 and ROCm 10
This release quietly cements llama.cpp as the universal inference runtime by adding default builds for CUDA 13.5 and ROCm 10.0 across Linux and Windows. The inclusion of ROCm 10 means AMD GPU users no longer need custom compilation to access modern acceleration, effectively closing the gap with NVIDIA's ecosystem support. While KleidiAI on Apple Silicon is disabled in this build, the broader platform coverage ensures that local inference remains accessible on virtually any hardware configuration without manual setup.
llama.cpp Releases