SolCompSolComp

~ decentralized compute

GPU network for AI inference on Solana

Run open-source LLMs (Llama, Mistral, etc.) at low cost. OpenAI-compatible API. Pay with SOL or USDC per token.

Get API KeyRead Docs
$ curl
curl https://api.solcomp.xyz/v1/chat/completions \
  -H "Authorization: Bearer sk_..." \
  -H "Content-Type: application/json" \
  -d '{"model": "llama-3.1-8b", "messages": [{"role": "user", "content": "Hello"}]}'
githubtwitter
compute

1.2M TFLOPS

nodes

12,450 online

requests

84.5M total

testnet

Real-Time Network Topology

Live connection status of global GPU nodes, different colors represent different GPU models

LIVE
GPU Model Legend
Tier 1
RTX 4090
RTX 4080 Super
RTX 4080
Tier 2
RTX 3090
RTX 3080 Ti
RTX 3080
Tier 3
RTX 3060 Ti
RTX 3060
GTX 1080 Ti
Active Nodes
0
Connections
0
Total TFLOPS
0.0K

## network

0.00M TFLOPS
live

Inference Request Data Flow

Watch in real-time how AI inference requests flow and are processed across the global GPU network

API Request Entry
Request Queue
0
84,523,456 total processed
Data Transfer
GPU Node Network
Active Nodes
12,450
Global Distributed Compute
Live Inference Stream
--:--:--

Global Node Distribution

12,450 GPU nodes forming a distributed neural network, providing low-latency compute for AI inference

NEURAL NETWORK LIVE
Active Connections
2.4M
Signals/sec
156K
Input
Hidden
Output
US-WNorth America West
27%
3,420 nodesAvg latency 32ms
US-ENorth America East
18%
2,180 nodesAvg latency 28ms
EUEurope
23%
2,850 nodesAvg latency 35ms
APACAsia Pacific
22%
2,780 nodesAvg latency 45ms
Global Total
12,450
Active GPU Nodes

## why solcomp

# problem

  • -High AI inference costs from cloud providers
  • -Millions of idle gaming GPUs worldwide
  • -Centralized compute monopolies

# solution

  • +Low-latency edge inference network
  • +Solana high-concurrency pooled compute
  • +Sub-second micro-settlements via cNFT

## architecture

Auto Benchmarking

Tier 1-3 GPU classification. Any hardware class supported.

Geo-Routing

<50ms latency. Smart task distribution across global network.

ZK Verification

Cryptographic proof of accurate inference. Trust-minimized.

## hardware tiers

tierhardwareuse caseest. earnings
Tier 1RTX 4090, H100, M-Ultra70B models$12-18/day
Tier 2RTX 4080, A100, M-Max7B-8B models$8-12/day
Tier 3RTX 4070, GTX 1080embeddings$5-8/day
latency: <50msaudit: 2% zk randompayout: daily solmin uptime: 95%

## products

Inference API

OpenAI-compatible. Run Llama, Mistral at low cost.

[api]

Connect GPU

Earn SOL by sharing idle compute power.

[earn]

Explorer

Real-time network stats and node map.

[dashboard]

## roadmap

Q1 2025Ignition[active]
  • -Settlement contract audit
  • -Genesis miner launch
  • -5,000 genesis nodes
Join Genesis Program →
Q2 2025Grid Sync
  • -ZK verification system
  • -Web3 API ecosystem
  • -B2B integrations
Q3 2025Power Grid
  • -Model sharding
  • -Community governance
  • -Cross-chain interop