Back to Organizations
Pluralis Research
Pluralis Research is a research lab developing decentralized AI through Protocol Learning, a communication-efficient approach to collaborative model training. It operates open training systems that allow distributed contributors to provide compute for collectively owned foundation models.
Distributed
Node0
Project Links
Description
Node0 is a decentralized, model-parallel AI training run powered by Pluralis Research’s Protocol Learning approach. Participants contribute GPU compute, each holding a portion of the model computation graph, while a dynamic network coordinates training across geographically distributed nodes. The official repository describes public participation with commodity hardware, including GPUs with at least 16 GB of memory, and provides installation and operation instructions.Category: Compute Network
Agora
Project Links
Description
Agora uses Protocol Learning, pipeline-parallel model sharding, and fault-tolerant collective operations so contributors can train large models over heterogeneous internet-connected GPUs without any participant holding the complete model.Category: Compute Network
Stoa
Project Links
Description
stoa separates reinforcement-learning post-training into rollout and training planes. Apple-silicon Macs generate rollouts with MLX, while CUDA GPUs train through Megatron/slime; both sides communicate through Cloudflare R2. Workers can join, leave, or crash during a run, while the trainer continues processing available rollouts.Category: Compute Network