Models

Prices are refreshed daily from OpenRouter. ExtraOrbital adds no model-price markup.

Every model

In / M and Out / M are USD per million input and output tokens. Free models may have provider rate limits. — means no published price; these models are excluded from the small tier.

Sort by clicking a column. The filter matches the model id and its display name.

Try one

Select a model and SDK to generate an example:

import OpenAI from "openai";

// Both variables are in .env already — the SDK reads them itself.
const client = new OpenAI();

const response = await client.chat.completions.create({
  model: "anthropic/claude-opus-5",
  messages: [{ role: "user", content: "Draft a status report from these logs." }],
});

console.log(response.choices[0]?.message.content);

Both OpenAI-compatible and Anthropic-compatible clients can call the catalog. Each example names the environment variables it needs.

Reading it as an agent

The same catalogue is available as JSON, without the page:

curl https://extraorbital.dev/api/gateway/models

{
  "models": [
    {
      "id": "anthropic/claude-opus-5",
      "name": "Anthropic: Claude Opus 5",
      "author": "anthropic",
      "contextLength": 200000,
      "promptPerM": 5,
      "completionPerM": 25,
      "modalities": ["text", "image"],
      "reasoning": true
    }
  ],
  "count": 420,
  "fetchedAt": "2026-09-01T00:00:00.000Z",
  "stale": false
}

stale: true means a stored snapshot was returned because the live fetch failed. Availability and prices may have changed.

The small tier

The model tier follows the team's plan: on the free Prototype plan every key is on the small tier, and on a paid plan every key reaches every model. There is no per-key setting. The small tier is set by price: every model priced up to $1.00 per million prompt tokens. Three speech-to-text models are included whatever their listed price, which is quoted per second or per audio token and does not compare with a text model's: openai/whisper-1, openai/gpt-4o-mini-transcribe and openai/gpt-4o-transcribe. A minute of audio costs under a cent on any of them.

The allowed model list is resolved from the catalog and refreshed every hour, so a key already provisioned picks up a model the tier gains, and every key of a team moves to the full catalog when the team leaves the free plan.

Prototype keys use this tier and spend from the shared $5 allowance. Paid plans have no model restriction.

A model outside the tier answers 404 with Model blocked by guardrail.

Next

• SDK recipes — every client in full
• Providers and keys — which SDKs reach which models