Cysic is putting more emphasis on running AI models. In a Nov. 3 interview with The Defiant, co-founder and chief executive Leo Fan said Cysic had seen a drop in demand for proofs from networks built on Ethereum.
Some networks built on Ethereum process transactions in batches and send Ethereum a summary. A piece of mathematical evidence called a zero-knowledge proof lets Ethereum check that the batch followed the rules without redoing every transaction. Cysic runs a network where people and companies can use their own hardware to generate proofs and earn digital tokens.
Crypto Briefing, citing Fan, reported that Cysic plans to run language models on graphics chips from that network. The job is called inference. It means running a trained model to do a task such as answering a question. The company introduced software for automating cryptocurrency tasks in March and still lists proof generation as a service.
Cysic’s developer documentation lists DeepSeek models under its inference service. It also lists models it provides through outside services. Cysic’s public materials do not say how much of its network hardware now runs AI models or how many customers pay to use it.
