Skip to content
OptimAI Router
Open Router
A NETWORK FOR OPEN MODELS

Open models.
One intelligent
connection.

Build with open models through Router API.
Serve models from your computer with Core Node.

Use Router API Explore Core Node
50+ open-model familiesMeta, DeepSeek, Qwen, Google, Mistral and moreFollow a request
POWERED BY OPTIMAI NETWORKROUTING, MADE VISIBLE
01 START WITH A REQUEST

Your request. Your model.
Your answer.

Pick the computer that serves your model, then follow the request through OptimAI Router and back to your application.

SCROLL TO TRACE THE PATH

Each computer serves one model. Click a room to choose.Choose a model:

YOUR APPLICATION
MODEL Qwen
Choose the computer, and the model it serves, for this routing illustration
Interactive illustration · Model names are examples, not live availability.
02 THE CONNECTION

You choose the model.
Router finds a computer.

Your request carries the model you selected. Router connects it to an available provider running that model.

Independent computers.
One familiar interface.

SAME MODEL. A CONNECTED NETWORK.

OPTIMAI ROUTEREXAMPLE REQUEST
READY TO CONNECTQWEN
MODEL-AWARE ROUTING

Real computers.
Run by people.

Models run on participating computers. OptimAI Router connects your application to the network, then carries the answer back.

Explore the playground A visual metaphor for software. Not a physical product.
03 THE RESPONSE

A request becomes
something useful.

Follow the signal to the result.
Text is demonstrated here. Image and video
cards explore the future visual direction.

Qwen text

ILLUSTRATION
Qwen/ EXAMPLE RESPONSE
Illustrative response: You choose the model. Router connects your request to an available provider and streams the answer back to your application.
RESPONSE PREVIEW01

Image generation

CONCEPT
FORMAT EXPLORATION · NOT LIVE02

Video generation

CONCEPT
FORMAT EXPLORATION · NOT LIVE03

Image and video are visual concepts, not a claim of current support. A selected text model does not generate all three formats.

FOR THE THINGS YOU’RE BUILDING
01

YOUR MODEL, CONNECTED

Keep your model choice. Reach an available provider through one API.

02

A NETWORK OF COMPUTERS

Open models, served by independent participating computers.

03

ANSWERS AS THEY ARRIVE

Stream the response into your app, or request a complete answer.

OptimAI Router
ONE CONTINUOUS PATH.
IMAGE PREVIEW

Mint horizon

Mint-green aurora sweeping above snow-covered mountain peaks and a still alpine lake
AI-generated concept artwork

The open AI economy

AI infrastructure should be
open on both sides.

OptimAI Router connects AI demand with a global, decentralized supply of models and compute.

01 / DEMANDIdeas become applications.
02 / SUPPLYCompute becomes possibility.
OptimAI Router

Independent participants. One open connection.

Hover, focus, or tap an object to explore.
Lower costsMore competitionBetter accessOpen participation

The ecosystem network loop

A virtuous cycle for a
more open AI economy.

Every new participant adds possibility. Each connection gives the next one a reason to grow.

01 / Participation

New providers connect their computers, adding resources to the network.

Model spectrum

Open models, called by name.

Each line of light is one edition in the Router catalog, ordered by size from 135M to 671B parameters and as tall as its context window. Put any edition’s id in your request’s model field.

100+open-model editions in the Router catalog

View all models

132 editions in the Router catalog

DeepSeek R1 8B

DeepSeek · 8B parameters · 128K context

  • Chat
  • Reasoning
deepseek-r1:8b

Try DeepSeek R1 8B in PlaygroundBecome a provider

Parameters 135M 671B

Taller = longer context

  • Reasoning
  • Coding
  • Chat
  • Size unlisted

Specs from the Router model catalog; context sizes are approximate. The API accepts text chat today, and live availability is on the Models page.

Each line is one of the 132 editions in the Router catalog, ordered by parameter size from 135M to 671B, followed by 7 editions whose catalog entry lists no size. Taller lines have longer context windows, from 2K to 1M tokens. Use the arrow keys to choose an edition; its details and actions follow.

The developer experience

From zero to your first request.

Connect in minutes.
Build something that matters.

One API

Switch models, not your stack.

A request is a stack of layers. Choose another model and only the top one moves; the endpoint, key, headers, and messages stay locked in place.

Example models

Example answerIllustrative, not a live call

Hello! It’s nice to meet you. How can I help today?

choices[0].message.contentqwen3:8b

  1. Model · the only change"model": "qwen3:8b"
  2. Messages“Say hello in one sentence.”Stays the same
  3. HeadersIdempotency-Key · X-Max-Charge-CreditsStays the same
  4. AuthorizationBearer $ROUTER_API_KEYStays the same
  5. EndpointPOST /v1/chat/completionsStays the same

Pick a model. Only the model layer changes.

Available

Build on a familiar interface.

Use the familiar chat completions API with cURL, Python, or TypeScript. Start with the example above.

Get your API key

Built for scale

From your first request to millions.

OptimAI Router scales with you. From prototypes to production, routing, infrastructure and billing continue through one system.

One system.

  • Routing
  • Infrastructure
  • Billing
Illustrative request flow

Made for the things you’re building

Infrastructure for
whatever you’re building.

Different ideas. One open foundation.

Find your model

AI Agents

Autonomous applications and workflows.

Search & RAG

Connect models to knowledge and archives.

Creative AI

Image, video and generative experiences.

Multimodal

Text, image, video, audio and vision.

Coding

Models for development workflows.

Research

Access models and distributed compute.

AI Agents, 1 of 6

Your idea. Your stack. One API.

For AI providers

Run a model.
Earn from demand.

Start with Core Node. Choose a model, keep it running, and let Router connect it to matching requests.

Try the app. Get a feel for what’s next.
Builder demandRouterYour computerHow it connects
OptimAI Core NodeInteractive demo · Simulated data

A little compute. A world of possibility.

Your node helps power the OptimAI Router — an open AI network, built together.

Model needed

Choose and prepare a model before serving.

Points

0.00

0.00· 24h
Credits earned

0.00

0.00· 24h
Requests completed

0

0· 24h
Your modelChoose a model
GPU: Example GPU (example)GPUExample GPUNetwork: 1 Gbps link (example)Network1 Gbps linkMemory: 16 GiB (example)Memory16 GiBCPU: 10 cores (example)CPU10 coresYour node

Your model

Choose a model to serve

Download a model, then select it for this computer.

Example computer

Your asking rates

Input
—
Output
—

USD / 1M tokens

Set rates to receive requests.

Recent tasks

On this computer · Example data

No requests yet

Requests appear here when your model answers them.

Run one model at a time.How serving works

Start serving in the banner to connect your selected model. Stop serving ends active work and releases its memory; the download stays saved. In the app, serving stops when your computer sleeps, you sign out, or you close it.

Tasks shown here are simulated. In the app, credits earned come from confirmed receipts; a completed task may still be waiting for its receipt.

Example computerLocal demo · no downloads or live requests

01 — CORE NODE / ROUTERInteractive demo. Simulated requests and credits.

Your computer can be someone’s next breakthrough.Run Core Node. Serve a model. Help build a more open AI economy.

Get Core Node (opens in a new tab)

Credit amounts are illustrative. Actual earnings depend on saved pricing and usage for paid requests. This demo awards no credits.