Idle GPUs In.
Lower Prices Out.
Cut your inference costs by up to 85% using our global network of idle GPUs. Compatible with OpenAI SDKs. No KYC required.
Hobbyist
Forever Free Tier
- 50,000 Free Tokens replenished weekly (P2P Models)
- $0.015 / 1M Tokens for Llama-3-8B & Qwen-7B
- Optional: $50/Year Unlimited Pass for 7B/8B models
- Community Support
Pro Builder
High Availability Swarm
- 5 Million Tokens included (Valid 90 days)
- Debug Mode: Free 100 tokens/min for dev
- Smart Routing: P2P & Data Center auto-switch
- Full Model Library Access (GPT-4o, Claude 3.5)
Enterprise
Global Scale API
- Volume Discount: 20% off for >100M tokens/mo
- Dedicated Instances: QPS > 5,000 guaranteed
- Private Deployment: Fine-tuning options
- 99.9% Uptime SLA & Account Manager
Model Registry
Simple pricing. No bundles. No tricks.
| Architecture | Provider | Context | Input Price | Output Price | Registry ID |
|---|---|---|---|---|---|
OpenAI/GPT-OSS-20B (Free)Free | P2P Cluster | 128k | $0.00 Per 1M tokens | $0.00 Per 1M tokens | |
Google/Gemini-2.0-Flash-Exp (Free)Free | P2P Cluster | 128k | $0.00 Per 1M tokens | $0.00 Per 1M tokens | |
Google/Gemma-3-27B-IT (Free)Free | P2P Cluster | 128k | $0.00 Per 1M tokens | $0.00 Per 1M tokens | |
DeepSeek/DeepSeek-V3.2Free | P2P Cluster | 128k | $0.00 Per 1M tokens | $0.00 Per 1M tokens | |
Meta/Llama-3.3-70B-Instruct (Free)Free | P2P Cluster | 128k | $0.00 Per 1M tokens | $0.00 Per 1M tokens | |
Qwen/Qwen3-235B-A22B (Free)Free | P2P Cluster | 128k | $0.00 Per 1M tokens | $0.00 Per 1M tokens | |
Amazon/Nova-2-Lite-V1 (Free)Free | P2P Cluster | 128k | $0.00 Per 1M tokens | $0.00 Per 1M tokens | |
TNG/DeepSeek-R1T-Chimera (Free)Free | P2P Cluster | 128k | $0.00 Per 1M tokens | $0.00 Per 1M tokens | |
TNG/DeepSeek-R1T2-Chimera (Free)Free | P2P Cluster | 128k | $0.00 Per 1M tokens | $0.00 Per 1M tokens | |
OpenAI/GPT-5.2-Chat | Official | 128k | $15.00▼ 33% Per 1M tokens | $60.00 Per 1M tokens |
Why Build with HETU?
1. Unbeatable Economics
By utilizing idle P2P resources (sunk costs) for small models, we drive prices down to near-zero. You save money; miners earn yield.
2. Developer-First Experience
Drop-in replacement for OpenAI. Just change your base_url and api_key. Includes a generous debugging allowance so you don't pay for errors.
3. Privacy & Freedom
We support USDT/USDC native payments. No intrusive KYC for standard tiers. Access uncensored models via our decentralized routing layer.
4. Hybrid Reliability
Best of both worlds: Ultra-cheap P2P nodes for background tasks, and premium Data Center nodes for production-critical workloads.
One Key. Real Use Cases.
For Developers
Drop-in OpenAI compatibility. Works with Python, Node.js, LangChain, and most OpenAI SDKs. Just change the base_url. Nothing else breaks.
client = OpenAI( base_url="https://api.hetuverse.com/subnet/api/v1/chat/completions", api_key="sk-..." )
For Power Users
Works with popular open-source clients. Use HETU directly in tools you already know: NextChat · LobeChat · Cherry Studio · Immersive Translate. Paste the key and endpoint. Done.
For RP & Privacy Users
Uncensored models, no moral lectures. Tired of filtered outputs and content warnings? Run Dolphin, Pygmalion, and other uncensored models with zero prompt logging.
Start in under 1 minute
Login with Wallet
No email or phone needed. Simply sign a message to login. Supports MetaMask, Phantom, OKX Wallet.
Recharge & Get API Key
Deposit USDC to activate your account. Then issue a personal token to access the network.
Change Base URL
Replace the official api.openai.com with our decentralized node address.
from openai import OpenAI
# Initialize with HETU config
client = OpenAI(
base_url="https://api.hetuverse.com/subnet/api/v1/chat/completions", # <--- Change this line
api_key="sk-hetu-your-key" # Paste your key here
)
response = client.chat.completions.create(
model="llama3-70b",
messages=[{"role": "user", "content": "Hello HETU!"}]
)
print(response.choices[0].message.content)FAQ
How can you offer Llama-3 at $0.015?
We utilize a global network of idle consumer GPUs (like RTX 4090s). Since the hardware cost is already "sunk" for the owners, we only need to cover electricity and a small incentive, passing huge savings to you.
Is the P2P network stable enough for production?
For the "Hobbyist" tier, we offer best-effort redundancy. For "Pro" and "Enterprise" tiers, our Smart Router automatically fails over to high-availability data center nodes if a P2P node lags, ensuring stability.
What is the "Free Debugging Allowance"?
We believe developers shouldn't pay for print("hello world"). Pro users get a allowance of 100 tokens/minute on specific small models strictly for testing and integration purposes.
Do you support Crypto payments?
Yes. We accept USDT (TRC20/ERC20), USDC, and ETH directly. No credit card is required for the Prepaid and Pay-as-you-go plans.