# OpenGPU Network — Full Technical Reference

> Extended llms.txt for AI agents and LLMs. Contains comprehensive technical documentation, API endpoints, protocol details, token economics, and integration guides for the OpenGPU decentralized GPU network.

## What is OpenGPU Network?

OpenGPU Network is a decentralized GPU compute network that connects GPU providers worldwide into a unified, on-demand compute marketplace. It offers AI training, inference, rendering, and high-performance workloads at up to 70% lower cost than centralized cloud providers (AWS, GCP, Azure). The network operates its own EVM-compatible Layer 1 blockchain (OpenGPU Chain) for transparent settlement, provider verification, and governance.

**Key value propositions:**

- Distributed GPU supply aggregated from thousands of providers globally
- Automatic workload routing and failover via the Relay Gateway
- EVM-compatible Layer 1 blockchain for on-chain settlement
- Proof-of-Execution consensus for verifiable compute
- Enterprise-grade SLA routing
- CertIK security audited

## Architecture Overview

### Layers

1. **Application Layer** — Client dApp, Provider Suite, dashboards
2. **Relay Gateway** — Enterprise routing, load balancing, SLA management
3. **Protocol Layer** — Smart contracts for task creation, matching, verification, rewards
4. **Blockchain Layer** — OpenGPU Chain (EVM L1) for settlement and governance
5. **Provider Network** — Global GPU nodes running the Provider Suite

### Core Components

- **Workload Router**: Intelligent global GPU routing with automatic load balancing and failover. Matches tasks to optimal providers based on latency, price, GPU type, and availability.
- **Relay Gateway**: Enterprise-grade routing layer. Fiat on-ramp, SLA guarantees, dedicated support. Live at https://relaygpu.com
- **OpenGPU Chain**: EVM-compatible Layer 1 blockchain using Lachesis DAG consensus. Handles payments, staking, governance, provider reputation.
- **Provider Suite**: Cross-platform desktop app (Windows, macOS, Linux) for GPU providers. Latest version: 3.1.0.
- **Client dApp**: Web-based interface for workload submission at https://client.opengpu.network
- **Management Dashboard**: Provider and fleet monitoring at https://management.opengpu.network

## Consensus & Verification

- **Consensus**: Lachesis DAG (Directed Acyclic Graph) — high throughput, low finality (~1s)
- **Proof-of-Execution**: Cryptographic proof that a GPU provider actually executed the workload
- **Provider Reputation**: On-chain reputation scoring based on historical execution accuracy, uptime, and task completion rates
- **Computation Consensus**: Redundant execution and result comparison for critical workloads

## Token Economics — OGPU

- **Name**: OpenGPU
- **Symbol**: OGPU
- **Type**: Utility token (ERC-20 compatible on OpenGPU Chain)
- **Decimals**: 18
- **Use Cases**:
  - Network settlement for GPU compute payments
  - Provider staking for reputation and rewards eligibility
  - Governance voting on protocol upgrades
  - Dynamic pricing bids for premium/priority workloads
- **Explorers**:
  - Mainnet: https://ogpuscan.io
  - Testnet: https://testnet.ogpuscan.io

### Exchanges & Listings

- MEXC: https://www.mexc.com/price/OGPU
- Gate.io: https://www.gate.io/trade/OGPU_USDT
- CoinMarketCap: https://coinmarketcap.com/currencies/open-gpu/
- CoinGecko: https://www.coingecko.com/en/coins/open-gpu

### Security Audit

- CertIK: https://skynet.certik.com/projects/opengpu

## Supported Workloads

- **AI Inference**: LLM inference, image generation, embeddings
- **AI Training**: Fine-tuning, full model training, RLHF, distillation
- **Batch Processing**: Dataset processing, feature extraction, preprocessing
- **Video Transcoding**: Hardware-accelerated video encoding/decoding
- **Rendering & VFX**: 3D rendering, animation frames, ray tracing
- **HPC Workloads**: Scientific computing, simulations, parallel compute

## Pricing

OpenGPU uses **dynamic pricing** based on:

- GPU type (consumer vs enterprise — RTX 4090, A100, H100, etc.)
- Current network demand
- Workload duration and priority
- Provider reputation and SLA tier

Typical savings: **50-70% lower** than AWS, Google Cloud, Azure for equivalent GPU types.

## Provider Suite — Download & Install

### Windows

- URL: https://opengpu.network/download/windows
- Format: `.exe` standalone installer
- Requirements: Windows 10/11, NVIDIA GPU with CUDA 11.0+, 8GB RAM minimum

### macOS

- URL: https://opengpu.network/download/macos
- Format: `.pkg` package installer
- Requirements: macOS 11.0+, Apple Silicon (M1/M2/M3) or Intel with Metal support

### Linux

- URL: https://opengpu.network/download/linux
- Format: Docker Compose
- Requirements: Ubuntu 20.04+, NVIDIA Container Toolkit, Docker 24+

### Provider Requirements

- Supported GPUs: NVIDIA RTX 3xxx/4xxx series, A100, H100, L40, Apple Silicon
- Minimum VRAM: 8GB for inference, 24GB recommended for training
- Internet: Stable connection, 100 Mbps+ recommended
- Wallet: EVM-compatible wallet to receive OGPU rewards

## Key URLs

### Main Pages

- Homepage: https://opengpu.network
- How OpenGPU Works: https://opengpu.network/how-ogpu-works
- Why OpenGPU: https://opengpu.network/why-ogpu
- Protocol: https://opengpu.network/protocol
- About: https://opengpu.network/about
- Get Started: https://opengpu.network/get-started
- How to Buy: https://opengpu.network/how-to-buy
- FAQ: https://opengpu.network/faq

### Solutions

- AI for Companies: https://opengpu.network/solutions/ai/companies
- AI Agents: https://opengpu.network/solutions/ai/agents
- AI Workloads: https://opengpu.network/solutions/ai/workloads
- GPU Providers: https://opengpu.network/solutions/infrastructure/providers
- Relay Gateway: https://opengpu.network/solutions/infrastructure/relay
- Enterprise: https://opengpu.network/solutions/infrastructure/enterprise
- Enterprise Pilot: https://opengpu.network/solutions/infrastructure/enterprise/pilot
- Blockchain: https://opengpu.network/solutions/blockchain
- Faucet: https://opengpu.network/solutions/blockchain/faucet

### Industries

- AI Training: https://opengpu.network/industries/ai-training
- Rendering & VFX: https://opengpu.network/industries/rendering-and-vfx

### Content

- Blog: https://opengpu.network/blog
- Glossary: https://opengpu.network/glossary
- Ecosystem: https://opengpu.network/ecosystem
- Careers: https://opengpu.network/careers
- Contact: https://opengpu.network/contact
- Media Kit: https://opengpu.network/media-kit

### Documents

- Litepaper: https://opengpu.network/docs/litepaper.pdf
- Whitepaper: https://opengpu.network/docs/whitepaper.pdf
- Privacy Policy: https://opengpu.network/legals/privacy-policy
- User Agreement: https://opengpu.network/legals/user-agreement
- Cookies Policy: https://opengpu.network/legals/cookies-policy
- Disclaimer: https://opengpu.network/legals/disclaimer

### External

- Developer Docs: https://docs.opengpu.network/
- Academy: https://academy.opengpu.network/
- Mainnet Explorer: https://ogpuscan.io
- Testnet Explorer: https://testnet.ogpuscan.io
- Client dApp: https://client.opengpu.network
- Management Dashboard: https://management.opengpu.network
- Relay Gateway: https://relaygpu.com

## Glossary of Key Terms

- **GPU**: Graphics Processing Unit — specialized hardware for parallel compute
- **GPU Node**: A physical or virtual machine contributing GPU resources to OpenGPU
- **Provider**: An entity running the Provider Suite to offer GPU capacity
- **Relay**: Optimized routing layer for enterprise workloads (Relay Gateway)
- **LLM**: Large Language Model
- **AI Inference**: Running a trained AI model to generate predictions
- **AI Training**: Teaching an AI model using labeled data
- **RAG**: Retrieval-Augmented Generation — LLM pattern using external knowledge
- **Fine-tuning**: Adapting a pre-trained model to a specific task
- **Checkpoint**: Saved model state during or after training
- **Embeddings**: Dense vector representations of text, images, or other data
- **TTFT**: Time To First Token — latency metric for LLM streaming
- **Mainnet**: Production blockchain network
- **Testnet**: Test blockchain network for developers
- **EVM**: Ethereum Virtual Machine — smart contract runtime
- **Smart Contract**: Self-executing code deployed on a blockchain
- **Staking**: Locking tokens to earn rewards and secure the network
- **Bridge**: Cross-chain asset transfer mechanism
- **DePIN**: Decentralized Physical Infrastructure Network
- **Proof-of-Execution**: Cryptographic proof that compute work was performed
- **Lachesis DAG**: Consensus protocol used by OpenGPU Chain
- **Intelligent Routing**: ML-based workload-to-provider matching

See the full glossary at https://opengpu.network/glossary

## Community & Support

- Twitter/X: https://x.com/openGPUnetwork
- Discord: https://discord.com/invite/opengpu
- Telegram: https://t.me/opengpuportal
- LinkedIn: https://www.linkedin.com/company/opengpu
- Medium: https://medium.com/@ogpunetwork
- YouTube: https://www.youtube.com/@opengpunetwork
- Email: info@opengpu.network
- Enterprise Sales: https://opengpu.network/solutions/infrastructure/enterprise?sales=true
- Security Contact: security@opengpu.network

## Network Statistics (as of network launch, March 2025)

- Status: Live on mainnet since March 2025
- Total tasks processed: 1,000,000+
- Active GPU providers: 230+
- Supported GPU types: NVIDIA RTX series, A100, H100, L40, Apple Silicon
- Global presence: GPU providers across 40+ countries

## License & Brand

- Brand Assets: https://opengpu.network/media-kit/branding-and-materials
- Logo Usage: https://opengpu.network/media-kit/logo
- Name: "OpenGPU Network" or "OpenGPU" (never "Open GPU" or "openGPU")
- Token Symbol: "OGPU"