Build once.
Route anywhere.

Create an API key in the console, then call any enabled model through the OpenAI-compatible endpoint. Use GET /v1/models with the same key to discover available IDs.

curl https://api.instamodel.ai/v1/chat/completions \
  -H "Authorization: Bearer $INSTAMODEL_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "'"$INSTAMODEL_MODEL_ID"'",
    "messages": [{"role": "user", "content": "Hello"}],
    "stream": false
  }'

Prepaid project keys currently accept non-streaming text requests only. Requests with stream: true fail before provider dispatch.

A gateway model ID currently maps to one configured upstream deployment. Paid publisher aliases must additionally snapshot the immutable revision, price, and terms before they are activated.

List models through GET /v1/models. Catalog metadata distinguishes platform-verified results from publisher-reported claims.

Instamodel-specific options live under the namespaced routing object, so standard OpenAI clients remain compatible.

{
  "model": "publisher/model-name",
  "routing": {
    "strategy": "balanced",
    "region": "us",
    "zero_retention": true,
    "allow_fallback": true
  }
}

Adapters translate provider protocols. The gateway owns routing, request IDs, usage records, and canonical errors. Some operational routes support SSE internally, but prepaid project keys remain non-streaming in this MVP.

Streaming is not activated for prepaid project keys. Before any future activation, a provider must pass streaming, tool-call, Unicode, usage, cancellation, context-limit, and error-mapping conformance tests.

The browser flow submits to POST /api/publisher/models. Each submission creates a model, an immutable revision, a price version, and a commercial-terms version together. Its agreement digest commits the license identifier, artifact reference, runtime declaration, prices, royalty, and exact attestation version. Organizations are quota-limited; validation, deployment, and alias activation remain operator approval gates.

1. 01DraftMetadata, license, base model, and task claims
2. 02QuarantinedArtifact present but not trusted or executable
3. 03ValidatingIntegrity, tokenizer, license, safety, smoke generation
4. 04DeployingApproved host receives digest-scoped artifact access
5. 05CanaryHealth, latency, output, and metering checks
6. 06LiveEligible for routing under the published alias

Submit an immutable Hugging Face or S3-compatible artifact reference with its SHA-256 digest. Direct browser uploads are not accepted in this build: all /api/publisher/uploads routes fail closed until private Blob ingestion and an isolated scanning pipeline are deployed.

An operator must independently verify the digest, license, runtime compatibility, and smoke-generation evidence before a revision can be activated. Pickle, remote code, custom Python, custom kernels, and uploader-provided runtime images are outside the trust boundary.

Console API keys use prepaid credit. Instamodel reserves the maximum token cost before dispatch, requires consistent terminal usage, and finalizes the measured charge before returning a paid response. For activated post-trained models, the wallet charge, request, attempt, token usage, and balanced royalty journal post atomically. Missing pricing, unsupported billable features, incomplete usage, or an ambiguous financial result fails closed.

gross_usage_charge = measured tokens × versioned retail rates
provisional_owner_accrual = gross_usage_charge × royalty_bps / 10,000
host_cost_entry = measured_units × versioned_host_rate
platform_entry = gross_usage_charge - owner_accrual - host_cost

Inference expense is kept separate from the model owner's percentage, so a routing change cannot silently alter creator economics. Purchase-level refunds and disputes are not yet attributed to usage entries, so every publisher figure is provisional rather than cash-backed, net, payable, or payout-ready. Statement closing and cash payouts are disabled.

Errors use the OpenAI object shape and every response carries x-request-id. Preserve that ID when opening a support issue or reconciling a usage line.

{
  "error": {
    "message": "No eligible deployment satisfies region=eu.",
    "type": "routing_error",
    "code": "no_eligible_deployment",
    "param": "routing.region"
  }
}