SiMa.ai introduced Palette 3.0 for deploying multiple AI models on Modalix chips. It supports container-based configurations and validation using sensor simulation, focusing on combining language, vision and other models in edge systems. The company cites up to a 3.5-fold speedup from speculative decoding and an example running GPT-OSS 20B on a 16 GB module. These figures are not guarantees for every model or input length, so actual memory use, latency and performance under shared resources require separate evaluation.
Back to Tech
SiMa.ai Palette 3.0 targets multiple AI models on edge hardware
Container deployment and speed claims depend on the model and hardware conditions.