為什麼 Mistral 押注其 80 億參數模型能在機器人導航上擊敗雷射雷達

mistral introduced robostral navigate, its first robotics model, which uses an 8b parameter architecture to navigate complex environments using just a standard camera. the system hits a 79.4% success rate on the r2r-ce benchmark for unknown environments, outperforming traditional methods that rely on depth sensors or multi-camera arrays. mistral trained the software entirely in simulation across 6,000 virtual spaces, applying it to wheeled, legged, and flying robots. early reinforcement learning experiments have already pushed performance higher by 3.2 percentage points with no plateau in sight. mistral positions navigation as the foundation for universal robotics, suggesting that efficient ai models could replace costly hardware dependencies.