On May 19, 2025, Skymizer launched HyperThought, a next-generation AI accelerator IP designed to enhance real-time, multimodal, and agent-based AI capabilities.

HyperThought builds on the success of EdgeThought, which revolutionized on-device LLM inferencing in 2024, providing improved performance and scalability.

This new platform is specifically designed to meet the growing demands for real-time, multimodal, and agent-based AI applications.

Powered by Skymizer's Language Instruction Set Architecture (LISA) v3, HyperThought supports multimodal processing and the concurrent execution of multiple AI models.

HyperThought is optimized for efficiency in compute, memory, and bandwidth, making it suitable for a wide range of edge AI deployments, from smart devices to autonomous systems.

The new LPU IP within HyperThought serves as a fully integrable subsystem, allowing System on Chip (SoC) teams to design custom AI silicon tailored for specific applications.

William Wei, CMO of Skymizer, highlighted that HyperThought goes beyond simple token acceleration, enabling intelligent agents capable of reasoning and real-time adaptation.

Additionally, the platform offers native support for agentic AI workflows, further enhancing the capabilities of edge devices.