Skip to content
toinferenceALPHABecome a provider
DEVELOPER DOCUMENTATION

One marketplace.
Your choice of inference.

Compare a specific model across independent providers. The interface and provider applications are available in preview. API access, payments, routing, and the verification agent are not live yet.

How the marketplace works

  1. Choose a modality and search for a model or provider.
  2. Filter for context, performance, price, precision, capabilities, location, and trust.
  3. Compare up to three deployments. Check their benchmark profile and weight evidence.
  4. Open a deployment for runtime details and an API contract example.

The deployment is the unit of comparison.

Two offers with the same model name can use different weights, precision, context limits, or serving engines. A production deployment will pin a model revision, tokenizer, adapters, quantization, runtime, provider, pricing schedule, and benchmark profile.

A 128K context window includes both input and output. A provider’s effective limit can be lower than the model’s theoretical limit. Benchmark performance at the context length you actually need.

Text API contract preview

The planned gateway uses an OpenAI-compatible request format with explicit provider selection. This example documents the intended contract and will not execute today.

POST https://api.toinference.com/v1/chat/completions
Authorization: Bearer YOUR_API_KEY
Content-Type: application/json

{
  "model": "Qwen3-32B",
  "provider": "qwen32-northstar",
  "messages": [{ "role": "user", "content": "Hello!" }],
  "stream": true
}

Pricing units

WorkloadDisplay unitCompare alongside
TextUSD per million input / output tokensTokenizer, context and cache policy
ImageUSD per imageResolution, steps, model, adapters
AudioUSD per audio minuteTask, language and audio settings
VideoUSD per configured clipLength, resolution, fps and steps
EmbeddingsUSD per million input tokensDimensions and batching

Provider integration

Providers will install the toinference agent, select a supported model revision, configure capacity and pricing, and complete benchmark checks before an offer is eligible for live traffic. Software downloads will appear here when the signed agent is released.

Model rights and permitted uses must be reviewed for each revision. Inclusion in this sample catalog does not assert that a provider has commercial serving rights.

Apply for the provider preview