Ocean Network Launches 'Inference' AI Service With Hourly GPU Pricing
Summary
- Ocean Network said it has launched Inference, a service that allows AI models to remain continuously deployed in dedicated GPU environments.
- Inference shows estimated costs before a session begins, charges based on GPU usage time, and starts at $2.16 per hour for an H200 GPU.
- Payments are processed on-chain through a Base blockchain-based escrow contract, and users receive free credits equal to six hours of H200 GPU usage.
Forecast Trend Report by Period
Ocean Network, a decentralized GPU cloud platform, has launched Inference, a service that allows AI models to remain continuously deployed in dedicated GPU environments.
Cointelegraph reported on September 10 that Ocean Network officially unveiled the Inference service that day. Users can continuously deploy AI models and applications on dedicated GPU nodes accessible via HTTP. The company described the launch as its biggest feature expansion since the platform went live.
Ocean Network cited opaque pricing, limited model transparency and uncertainty over hardware allocation as key problems in the existing AI infrastructure market. Inference is designed to address those issues by showing estimated costs before a session starts and charging by GPU usage time rather than token consumption. Each request is also routed to a dedicated node rather than shared infrastructure.
Pricing starts at $2.16 an hour for an H200 GPU. In internal testing, the service generated a 30-second video in about 20 minutes at a cost of $0.72. Payments are processed on-chain through an escrow contract based on the Base blockchain.
The service is available in four formats: curated models, custom models, services and templates. Curated models include preconfigured options such as Qwen3-8B, while custom models support models listed on the Hugging Face Hub. To mark the launch, the company is offering free credits equal to six hours of H200 GPU usage.
