Open Network · Decentralized Hosting

The free market for AI inference.

Anyone can host models. Anyone can earn. No gatekeepers, no central committee — just hardware, performance scores, and open competition for routing rewards.

1,240+
Active model hosts
across 38 countries
380+
Models available
open and fine-tuned
94 ms
Avg routing latency
p50, global
$2.1M+
Tokens distributed
to node operators this year
How it works

Host. Compete. Earn.

Four steps from bare metal to token income. No permission needed — the protocol decides routing, not a person.

Step 01

Provision your hardware

Connect any GPU or CPU cluster to the Elis network. Set your own pricing, choose which model families you support, and go live in minutes. No central approval needed.

Step 02

Serve inference requests

The Elis routing layer sends requests to the best available provider based on latency, cost, and reputation score. Your node competes on merit — the market decides.

Step 03

Earn based on performance

Token credits are weighted by output quality scores, uptime, and response latency. Better performance means a bigger share of each routing epoch's credit pool.

Step 04

Put them to work

Spend $ELIS credits within the network — stake them to boost your routing priority, delegate to a pool, or put them toward premium model access. Credits stay in-network.

Host one model, access them all.

Contributing a node to the network gives you routing access to every other model on it — from compact 7B models to frontier 70B+ checkpoints hosted by other community operators. Run the hardware you have, use the models you need.

  • 380+ open and fine-tuned models available immediately
  • Switch models mid-workflow without re-provisioning
  • Access specialized fine-tunes: legal, medical, code, multilingual
  • Your node earns while you route traffic elsewhere

A smarter MCP server that preserves tokens.

Elis ships a token-aware MCP server that routes each tool call and context window intelligently — trimming redundant content, caching repeated context, and picking the cheapest capable model for each subtask. You get the same output quality at a fraction of the token spend.

Context compression
Strips repeated boilerplate from multi-turn tool calls before they hit the model.
Tiered model routing
Simple subtasks go to small fast models. Only complex reasoning reaches the big ones.
Shared context cache
System prompts and tool schemas cached across requests — not re-encoded every time.
Token Economy

$ELIS — performance pays.

Tokens are not handed out on a schedule. They flow toward the nodes that actually deliver — creating a self-correcting market that rewards quality over presence.

Performance-weighted rewards

Every epoch, the reward pool is distributed proportionally to quality scores. Low-latency, high-accuracy nodes earn more — no fixed salary, pure market signal.

Transparent on-chain accounting

Every payment, every quality score, every routing decision is logged on-chain. Anyone can audit the ledger — no black boxes, no hidden fees.

Stake to boost routing priority

Stake $ELIS tokens against your node to signal commitment. Higher stake weight increases your share of incoming requests during high-demand windows.

Delegate to a mining pool

Don't want to run a node solo? Delegate your stake to a community mining pool, earn a proportional share of pool rewards, and let experienced operators handle the infra.

Mining Pools

Participate at any scale.

Whether you are spinning up a single GPU rig in your garage or running enterprise infrastructure, there is a path into the network.

Solo Miner

Any amount

Run your own node, keep 100% of your earnings. Full autonomy, full responsibility.

  • Custom pricing control
  • Direct performance scoring
  • Credit balance tracking
  • On-chain credit settlement every epoch
Most popular

Community Pool

Minimum 100 $ELIS

Join a managed pool. Shared infra, shared rewards, lower variance.

  • Pool operator handles uptime
  • Smoothed reward distribution
  • Priority routing allocation
  • Weekly pool leaderboard

Enterprise Node

10,000+ $ELIS

Deploy a dedicated high-throughput node for regulated industry traffic. SLA-backed.

  • Guaranteed minimum load
  • Compliance routing tags
  • Private pool option
  • Dedicated support channel
For regulated industries

Compliance without lock-in.

Regulated teams — legal, healthcare, finance, government — need model provenance, audit logs, and deployment control. The Elis community network delivers all three without handing sovereignty to a single cloud vendor.

  • Choose providers with HIPAA / SOC 2 attestations from the registry
  • On-chain routing logs satisfy evidence requirements for audits
  • Switch providers instantly if SLA or compliance posture changes
  • Run a private enterprise pool for fully internal inference
Network guarantees
Provider attestations
Nodes self-certify compliance posture. Enterprises can filter to only route to attested providers.
Immutable routing log
Every request is logged to an append-only on-chain ledger. Pull an audit trail on demand.
No single point of control
No Elis employee can reroute your traffic or suppress a node. Protocol rules are code.
Exit anytime
Your data never locks in. Withdraw stake, move traffic, or spin down with zero penalty.

Common questions

What hardware do I need to participate?
Any NVIDIA GPU (16 GB VRAM minimum recommended) or CPU cluster can participate. The network accepts providers of all sizes — a single RTX 4090 up to multi-node A100 clusters.
How are quality scores calculated?
The Elis scoring system runs blind evaluation prompts through your node periodically, compares outputs against ground-truth benchmarks and peer consensus, and produces a rolling quality score. Full methodology is published in the community docs.
Can I run models I fine-tuned myself?
Yes. You can register any HuggingFace-compatible model checkpoint. If your model performs well on community benchmarks it will receive traffic — the market rewards quality regardless of origin.
What is the $ELIS token?
$ELIS is the network's in-network utility credit. It is earned by serving inference requests and spent within the network on routing priority, model access, and staking. There is no external withdrawal — credits circulate inside the Elis network.
How do mining pools work?
A pool operator runs one or more nodes and accepts delegated stake from other holders. Rewards are split proportionally to each delegator's stake contribution minus a pool fee set by the operator. All splits are enforced on-chain.

Ready to own your compute?

Plug in your hardware, register your first node, and start earning $ELIS in the first epoch after you go live.