Acurast deploys open-source AI decision model Laya on decentralized smartphone network, achieving sub-second real-time inference

TechFlow: On September 28, decentralized computing network Acurast announced the large-scale deployment of the open-source "System 1" decision-making model Laya across its global smartphone nodes. Developed by Convai Innovations under the Apache-2.0 license, the model focuses less on text generation and instead evaluates options provided by applications to make real-time judgments and execute subsequent actions, making it suitable for scenarios such as game control, autonomous navigation, email classification, content recognition, and security auditing.

Acurast stated that every decision made by Laya is executed directly via the CPU on Android smartphone nodes, utilizing the device's built-in Trusted Execution Environment (TEE) to ensure computational confidentiality and verifiability, with each decision typically taking approximately 0.2 to 1 second. Currently, the Acurast network is connected to over 280,000 smartphones across more than 175 countries globally, having cumulatively processed over 918 million on-chain transactions.