AI Router models and pricing · Galaxium Hosting help

Read the Models page, work out what a request costs in credits, and pick the right model for the job.

AI Router → Models lists every model you can use, with its price. One API key and the same request format work for all of them.

Read the Models page

Each row shows:

• Model: the name, and the model ID you send in requests, for example openai/gpt-4o-mini. Click the copy icon next to it.
• Context: how much text (in tokens) the model can take in one request, including the conversation so far.
• Input / 1K and Output / 1K: the price in credits per 1,000 tokens you send and per 1,000 tokens the model writes back.
• Tags: hints such as cheap, fast, coding, reasoning, vision or long context.
• Your 30d: how many requests you sent to it in the last 30 days.

Filter by provider with the buttons at the top, search by name, ID or tag, and sort by Cheapest input, Cheapest output, Largest context or Most used by you.

Prices can change as model providers change theirs. The Models page always shows the current price.

What is a token?

Models count text in tokens. A token is a piece of a word: in English, 1,000 tokens is roughly 750 words. Both your prompt (input) and the answer (output) are counted.

Work out a cost

Cost in credits = (input tokens ÷ 1,000 × input price) + (output tokens ÷ 1,000 × output price).

The easy way: click Estimate on a model's row, enter Input tokens, Output tokens and the number of Requests, and the window shows the cost in credits and dollars.

Every API response also tells you what that request cost (see AI Router usage, balance and limits).

Save with the response cache

When you send exactly the same request again (same model, messages and settings) within the cache lifetime, the stored answer comes back instantly at 10% of the normal price. Turn it on or off, and set the lifetime, under Settings → Response caching. Cached answers are private to your account.

To skip the cache for one request, send the header X-Galaxium-Cache: bypass or "cache": false in the request body.

Pick a model

• Start cheap. For chat bots, summaries and simple tasks, try a model tagged cheap or fast first.
• Step up when needed. Use a reasoning or coding model for harder problems, maths or code.
• Long documents need a large Context.
• Test in the Playground with your real prompts before you switch your app over. Try on a model's row opens it there.

Set the model your requests use when they don't name one under Settings → Default model.