Ryvion is an OpenAI-compatible inference network with verified execution. Send your AI workloads to distributed GPU nodes worldwide and get cryptographic proof of every job.
- Chat completions -- LLM inference with streaming support (phi-4, Llama 3.2, TinyLlama BitNet)
- Embeddings -- text embeddings for search and RAG (nomic-embed-text-v1.5)
- Image generation -- SDXL Turbo and Stable Diffusion XL
- Audio transcription -- Whisper-based speech-to-text
- Knowledge bases -- upload documents, embed them, and query with semantic search
- RAG-powered chat -- attach knowledge bases to chat completions
- Agent memory -- persistent key-value memories and session diaries for AI agents
- Agent hosting -- deploy persistent agents on sovereign GPU nodes
- MCP integration -- connect Claude Code, CrewAI, or any MCP client directly
- OpenAI-compatible -- change
base_urlandapi_key, your existing code keeps working - Cryptographic receipts -- Ed25519 signed proof on every job (node, GPU, region, timestamps)
- Sovereign routing -- pin workloads to any of 9 verified jurisdictions
- Guardrails -- PII detection, prompt injection defense, content safety, full audit trail
- EU AI Act ready -- compliance-grade audit trail built into the protocol
https://ryvion-hub.fly.dev
- Create an account -- every new account gets $1 CAD in free credits
- Generate an API key
- Follow the Quick Start guide
| Surface | Price |
|---|---|
| Chat completions | $0.06 CAD / 1M tokens |
| CPU inference (BitNet) | $0.03 CAD / 1M tokens |
| Embeddings | $0.06 CAD / 1M tokens |
| Image generation | $0.03 CAD / image |
| Audio transcription | $0.006 CAD / minute |
| Knowledge base search | $0.01 CAD / query |
| Agent hosting | $0.50 CAD / hour |
| Fine-tuning | $2.00 CAD / run |