AI inference at the price of electricity
SeedInfer is a decentralized, peer-to-peer inference network. Your requests run on verified NVIDIA GPUs owned by independent providers — through an OpenAI-compatible API, at near-electricity cost.
- OpenAI-compatible
- Verified NVIDIA GPUs
- 99% to providers
Gemma 4 26B A4B NVFP4
google/gemma-4-26b-a4b-nvfp4
- Context
- 256K
- Input
- $0.03
- Output
- $0.20
USD per 1M tokens
- Cached input is free (60s TTL, max 5 min)
- NVFP4 weights on NVIDIA Blackwell
Live network
Network status, refreshed every 15 seconds
Developers
A drop-in OpenAI replacement
Change two lines — the base URL and the key. Chat Completions, including streaming, work with the SDKs you already use.
from openai import OpenAI
client = OpenAI(
base_url="https://seedinfer.com/v1",
api_key="sk-seedinfer-...YOUR_KEY",
)
resp = client.chat.completions.create(
model="google/gemma-4-26b-a4b-nvfp4",
messages=[{"role": "user", "content": "Hello from SeedInfer!"}],
)
print(resp.choices[0].message.content)How it works
From request to verified GPU in three steps
- 01
Create an account
Sign up and top up pay-as-you-go credit in USDC or ETH — no minimum commitment. API keys are in early access during the public beta.
- 02
Point your OpenAI client
Swap the base URL for https://seedinfer.com/v1. Existing SDKs, tools and agents work unchanged.
- 03
Verified GPUs serve it
Requests are routed to verified NVIDIA GPUs run by independent providers, priced close to the electricity they burn.
For GPU owners
Keep 99% of what your GPU earns
Got an NVIDIA card with 32 GB+ of VRAM? Install the agent with one command and start serving inference. The network takes a 1% fee — the rest is yours. Payouts: USDC on Base, monthly, minimum $1.00.