World models have mostly been too large to run on the robot itself. A four billion parameter variant with a published post-training path changes where the policy can live.

• articleEvaluating AI Agent Skill Performance with NVIDIA SkillEvaluator
• articleServe Qwen3.8-2.4T-A95B, a 2.4T-Parameter Model, with Configurable Reasoning on NVIDIA GB300 NVL72
• articleRoute AI Agent Workloads Across Models with NVIDIA NeMo Switchyard
• articleRun Local Agentic AI Workflows with Meta’s Muse Glimmer on NVIDIA

• articlePost-Train NVIDIA Cosmos 3 in One Day Using Agent SkillsTanya Lenz
• videoWhy Large? Tiny LMs & Agents on Edge/Robotics — Cormac Brick, GoogleAI Engineer
• videoLearning on the Job: The Future of Post-Training — Raymond Feng, Applied ComputeAI Engineer

Sign in to comment.

Loading comments…