Global physical AI company SiMa.ai said on Wednesday it has unveiled Palette 3.0, a major software update that supports execution across physical AI models.
The company said Palette 3.0 supports AI inference and actions in real-world environments through improved performance for running multiple models simultaneously. It also boosts model tokens-per-second throughput by up to 3.5 times in low-power environments.
It increases processing performance for perception, depth and inference models needed for drones and robotics. It also provides new open-source features for developing generative AI applications.
SiMa.ai explained that it worked with ecosystem partners to add sensor fusion and digital twin simulation features, easing integration and verification bottlenecks in developing autonomous systems.
Krishna Rangasai (크리슈나 랑가사이), CEO of SiMa.ai, said the fundamental bottleneck in physical AI today comes from a clear gap between software that evolves every month and legacy silicon hardware that takes years to respond. He said Palette 3.0 delivers SiMa.ai's core vision of “hardware that keeps pace with software progress.” He said it continues to maximize performance per watt across all physical AI models, helping customers implement future-ready capabilities on current hardware and fundamentally changing the roadmap for developing fully autonomous systems.