The open network for specialist models

Use the best-fit model forspecialized work across every industry.

One API for frontier models and the world's best post-trained specialists. Publish your model, connect compute, and define versioned terms for usage-based earnings.

Routing
Multi-provider
API format
OpenAI
Creator terms
Versioned
ILLUSTRATIVE ROUTEConfigured deployment
Example
INPUT / task: contract_review

Compare the limitation-of-liability language against our policy and return the material exceptions.

NC
SELECTED MODELnorthstar-demo/clauseguard-32bserved by Baseten · us-east
FIT98%
LATENCY312 ms
COST$0.0008
CREATOR SHAREmetered
01 / MODEL EXCHANGE

Specialists, ready to route.

Compare quality, speed, and price across base and post-trained models—with one contract and one API.

Northstar Legal · fictional demoDEMO CLAIM

clauseguard-32b

Legal reviewPost-trained
Eval
91.4 CUAD
Input
$0.32 / 1M
View model
Lattice Commerce · fictional demoDEMO CLAIM

catalog-match-8b

CommercePost-trained
Eval
97.1 F1
Input
$0.08 / 1M
View model
HelioHelp Labs · fictional demoDEMO CLAIM

support-resolve-14b

Customer supportPost-trained
Eval
86.7 Resolve
Input
$0.38 / 1M
View model
02 / ONE NETWORK, TWO SIDES

Built for callers.
Built for creators.

Model ownership, inference hosting, and API access stay separate—so each party gets one clear contract, one ledger, and the right controls.

A / FOR AI TEAMS

One key. The right model for every request.

Keep the OpenAI client you already use. Choose an exact model and constrain its configured deployment by region and retention policy.

  • OpenAI-compatible chat completions
  • Streaming, tools, and structured output
  • Fail-closed region and retention constraints
  • Per-request provider, latency, and usage receipts
quickstart.pyPYTHON
from openai import OpenAI

client = OpenAI(
  base_url="https://api.instamodel.ai/v1",
  api_key="mr_live_..."
)

response = client.chat.completions.create(
  model="northstar-demo/clauseguard-32b",
  messages=[{"role": "user", "content": prompt}]
)
Create an API key
B / FOR MODEL LABS

Your specialist model should earn while it works.

Bring a post-trained model or LoRA adapter. The private alpha records an immutable artifact, price, and royalty agreement while validation, deployment, and payout activation stay approval-gated.

  1. 01
    ImportSafetensors, Hugging Face, or S3-compatible storage
  2. 02
    ValidateProvenance, compatibility, safety, and smoke generation
  3. 03
    DeployImmutable version served by an approved inference host
  4. 04
    EarnAuditable usage ledger and monthly settlement
EXAMPLE REQUEST ECONOMICSUSD
Customer usage charge$1.00
Model owner royalty+$0.62
Inference host + platform$0.38
Illustrative only. Actual terms are versioned per model.
Start publishing
03 / TRUSTED CONTROL PLANE

Build toward auditable inference.

Publisher submissions already snapshot immutable revisions, prices, and commercial terms. Connecting those snapshots to every paid gateway request is an explicit release gate for the marketplace.

Read the routing contract
MODEL OWNER
Immutable weightsrevision sha256:8f2…
Commercial termsroyalty · price · license
INSTAMODEL
Gateway receiptrequest · provider · region
Ledger schematokens · charge · royalty
INFERENCE NETWORK
Approved hostcapacity · health · retention
Provider usagerequest · units · latency
01No silent alias changeThe requested public model ID is preserved on every response.
02Prompts stay privatePublishers never receive customer prompts or identities.
03One canonical requestIdempotency reserves a customer request before dispatch.
04Versioned economicsPublisher pricing and royalty terms are stored as immutable versions.
THE MODEL NETWORK AFTER PRETRAINING

One key. A marketplace
of models.

Start with the public deterministic demo—no account or API key required. Paid provider routes open after checkout and cash-backed usage accounting are connected.

Try public demo Publish a model