RELEASE · MODELS · #2933
NVIDIA releases Cosmos Framework with Cosmos 3 Nano simulation recipes and NIM microservices
NVIDIA announced the Cosmos Framework, a modular platform of world foundation models and developer tools for physical AI that supports training, RL, optimization and inference across models including Cosmos 3, Alpamayo, GR00T, Qwen and pi0.5. The release includes Cosmos 3 Nano simulation recipes and example checkpoints (PyTorch/vLLM-Omni), NIM microservices for accelerated inference (quantization, speculative decoding, KV-cache-aware routing) with ~2–3x speedups on H200 GPUs, and guidance on GPU memory use (Cosmos3-Nano: ~58 GiB BF16; ~44 GiB FP8 with offloading); a livestream is scheduled for Nov. 5.
KEY POINTS
- NVIDIA announced the Cosmos Framework, a modular platform of world foundation models and developer tools for physical AI that supports training, RL, optimization and inference across models including Cosmos 3, Alpamayo, GR00T, Qwen and pi0.5.
- The release includes Cosmos 3 Nano simulation recipes and example checkpoints (PyTorch/vLLM-Omni), NIM microservices for accelerated inference (quantization, speculative decoding, KV-cache-aware routing) with ~2–3x speedups on H200 GPUs, and guidance on GPU memory use (Cosmos3-Nano: ~58 GiB BF16; ~44 GiB FP8 with offloading); a livestream is scheduled for Nov.
- This matters because Cosmos combines world-model checkpoints, simulation recipes and GPU-optimized serving to help developers adapt and deploy embodied/physical AI systems more quickly and efficiently.
WHY IT MATTERS
This matters because Cosmos combines world-model checkpoints, simulation recipes and GPU-optimized serving to help developers adapt and deploy embodied/physical AI systems more quickly and efficiently.