Compute Layer

How are inference requests handled?

When a user sends a request to a model, the Model Service validates the user’s balance and forwards the job to the Node Service. Nodes execute the model through the Node Execution Engine and return the result. Payments and rating updates are processed automatically afterward.

How is workload distributed and scheduled?

The Node Service selects Nodes based on availability, performance, and rating. DeepNode uses a “One Model – Two Nodes” policy to balance load, provide redundancy, and rotate workloads across the network. This prevents overloading and gives all Nodes fair earning opportunities.

What hardware requirements exist for running Nodes?

Nodes must provide reliable compute resources. GPU-backed machines are recommended for heavier ML models, while CPUs may be sufficient for lightweight inference. Requirements vary by model and generally include:

• CPU Nodes for low-latency, lighter workloads.

• GPU Nodes (NVIDIA or AMD) for deep learning and LLM inference.

Miners are responsible for ensuring uptime, correct drivers, and proper resource allocation.

What are the performance expectations?

Performance targets vary by model type, but Miners should expect:

• Low latency for single request.

• High throughput for batch tasks.

• Stable uptime as tracked by the Node Service.

Performance directly affects a Node’s rating, influencing workload allocation and rewards.

How do I build reputation as a Node Operator or Validator?

Reputation in DeepNode grows through consistent, reliable participation:

• Node Operators build trust by maintaining uptime, handling inference requests correctly, and avoiding errors. The “One Model – Two Nodes” policy pairs new or smaller Nodes with high-reputation Nodes, giving them fair opportunities to prove themselves.

• Validators gain reputation by submitting accurate, timely, and independent validation results. Copied or biased decisions reduce reputation and may lead to penalties.

• Dynamic Trust Weights adjust influence over time, allowing newcomers to gradually increase their weight as they demonstrate reliability.

This system ensures a balance: established participants maintain quality, while new participants always have a path to earn trust and secure more opportunities.