Perceptron unveils a 36B robotics model

Perceptron unveiled Isaac 0.5, a 36 billion-parameter sparse robotics model designed to share one sequence across video, language, robot state, and action history. Its public model card describes evaluation across 35 embodiments and lists code, interfaces, benchmarks, manifests, and fine-tuning paths, while its open repository is Apache 2.0 and says the model weights are still coming soon. For robot teams, the useful signal is not a benchmark victory but a broader reproducibility package: compare it with simulation-first robotics screening before budgeting scarce hardware hours.