Virtuals Protocol has integrated Leyten’s distributed inference engine to deploy GLM-5.2 across its decentralized AI agent network. This strategic partnership enables the platform to run massive language models without relying on centralized cloud providers or single GPU clusters.
Leyten’s shard engine utilizes pipeline-parallel inference to slice large models and distribute them across multiple networked GPUs. This architecture ensures no single node must hold the entire model in memory, solving critical hardware bottlenecks for decentralized compute.
The integration targets GLM-5.2, an open-weight model from Z.ai released under an MIT license. Featuring approximately 744 billion total parameters and a one-million-token context window, the model uses a mixture-of-experts architecture to maintain manageable compute costs while delivering frontier-level performance.
This infrastructure upgrade directly supports Virtuals Protocol’s native VIRTUAL token ecosystem. By leveraging GLM-5.2’s advanced agentic coding capabilities and extended context, the platform enhances the autonomy and transactional efficiency of its onchain AI agents.