OpenRouter

Connect OpenRouter to reach hundreds of models through one key - with notes on credits, choosing models, and how token billing works.

OpenRouter is a gateway that exposes hundreds of models from many providers (OpenAI, Anthropic, Google, Meta, Mistral, DeepSeek, and more) through one API key and one balance. It is the easiest way to try many models without separate accounts, and to switch between them freely.

No subscription needed. OpenRouter is pure pay-as-you-go: you top up credit once and it routes each request to the underlying provider.

Step 1 - at OpenRouter

  1. Sign in at openrouter.ai.
  2. Go to openrouter.ai/keys, create a key, and copy it.
  3. Add credit at openrouter.ai/credits. OpenRouter is prepaid; a few dollars goes a long way. A small processing fee applies to top-ups (see their docs for current rates).

Step 2 - in Sciwand

  1. Open Settings -> AI Models, then click Manage (or Add Model) to open Configure Models.
  2. Select OpenRouter, then click Configure.
  3. Paste the key into API Key and click Save Changes.
  4. Tick the models you want to enable.

Understanding the models

  • Model names are written as vendor/model, for example openai/gpt-..., anthropic/claude-..., google/gemini-..., or meta-llama/llama-... (the part after the vendor includes the specific version).
  • Browse the full catalogue, with live prices and context lengths, at openrouter.ai/models.

Free models

OpenRouter offers a rotating set of free models (their names end in :free), refreshed regularly - roughly monthly - as new open models appear. To use them at usable rate limits you need to have added credit at least once: a one-time top-up of about $5 is the minimum, and the free models then keep working without drawing it down. That single charge also raises your daily limits, and the balance is there if you later want to try a paid model. See Free and low-cost models.

How billing is calculated

OpenRouter charges per token, at the rate set for each model (a separate price for input/prompt tokens and output/completion tokens, usually quoted per 1M tokens). The cost of one request is roughly:

(prompt tokens x input price) + (completion tokens x output price)

So longer documents (more input) and longer answers (more output) cost more, and a premium model can cost many times what a small one does for the same task. A few models also add a small per-request or per-image charge. Your remaining balance and a full usage log are shown in your OpenRouter dashboard.

Choosing a good model

  • Everyday questions and summaries: a small, fast model (look for -mini, -flash, or -lite in the name).
  • Hard reasoning or careful writing: a larger flagship model (the full-size, non-mini variants).
  • Long papers: check the model context length on the models page so the whole document fits.
  • Watch the per-model price shown next to each entry and match it to the task - you do not need the most expensive model for routine work.

More on OpenRouter

For provider-specific behaviour, rate limits, the auto router, and current fees, see the OpenRouter documentation. See also the general chat setup guide.