AI inference,
priced per token,
paid in SOL.
InferNode is a Solana-native marketplace where buyers submit AI jobs and independent providers earn by running models on idle GPUs and OpenAI-compatible endpoints.
$ infernode worker start
[14:02:11] auth ........ ok (provider_8x4f)
[14:02:11] endpoint .... http://localhost:11434
[14:02:11] models ...... llama3.1, mistral
[14:02:12] polling jobs every 2s ...
[14:02:14] job_a91b accepted
task: text-generation
model: llama3.1
input: 312 tok
[14:02:18] inference complete ✓
output: 248 tok | 4.1s
[14:02:18] submitting result hash ...
[14:02:19] payout queued 0.00284 SOLFrom prompt to payout in four steps.
Submit
Connect wallet. Choose a task — text generation, summarization, embeddings, classification. Pay into a Solana escrow.
POST /jobs
{ task: 'text-generation',
model: 'llama3.1',
input: '...' }Dispatch
The job lands in a queue. Capable, staked providers running infernode-worker poll for matching tasks.
worker.poll() → job_a91b llama3.1 · 312 tok
Execute
Provider runs inference locally via Ollama / vLLM or an OpenAI-compatible API. Result + hash submitted back.
POST /jobs/a91b/result
{ output: '...',
hash: '0x7Hk2…fQp9' }Settle
Anchor program releases payment to the provider, minus protocol fee. Refunds on timeout. Disputable window for failures.
release_payment(job_a91b) → provider: 0.00284 SOL → treasury: 0.00015 SOL
Cheap inference, on demand.
Submit jobs without API keys or vendor lock-in. Pay only for the tokens you use. Choose the model and price point that fits.
- Wallet-native: no signup
- Per-token pricing, transparent fees
- Multiple supported task types
- Job history + result hashes on-chain
Monetize idle compute.
Run infernode-worker on your GPU box, VPS, or wire up a hosted OpenAI-compatible endpoint. Stake to join, earn per job.
- Ollama, vLLM & OpenAI-compatible
- Set your own prices per model
- Provider stake + reputation
- Automatic payouts via Anchor escrow
Models available across the network.
Simple, deterministic pricing.
price = baseFee + (estimatedTokens / 1000) * pricePerKTokens fee = price * protocolFeePct payout = price - fee
- base0.00010
- tokens (2.5 × 0.0005)0.00125
- subtotal0.00135
- protocol fee (5%)-0.00007
- Provider receives0.00128 SOL
Hybrid by design. Trustless where it matters.
┌──────────────────┐ submit job + pay ┌──────────────────────┐
│ Buyer (web) │ ─────────────────────▶ │ Anchor Escrow PDA │
│ wallet · prompt │ │ amount · expires_at │
└────────┬─────────┘ └──────────┬───────────┘
│ │
│ job metadata │ assign · release
▼ ▼
┌──────────────────┐ dispatch via queue ┌──────────────────────┐
│ Backend (API) │ ───────────────────────▶│ Worker CLI (provider)│
│ postgres·redis │ ◀───────────────────────│ ollama / vllm / api │
└────────┬─────────┘ result + hash └──────────────────────┘
│
▼
buyer sees result, provider receives payoutProvider registry, escrow PDAs, payment release, refund + slash paths.
Job queue, dispatcher, result storage, reputation scoring.
Worker CLI on provider hardware. Polls jobs, runs inference, submits result + hash.
Compute should be a commodity.
Help us build it. Submit your first inference job in under a minute, or spin up a worker tonight.