Free and Low-Cost Models

You can use Sciwand for little or nothing by choosing free tiers, free models, or providers that only need a small one-time top-up. Here are the main options for both chat and embeddings - see Connecting AI Providers for the basics first.

Free online options

• Google Gemini - the Gemini API has a free tier for both chat and embeddings. A great first choice if you do not want to pay.
• Groq Cloud - a free tier with very fast responses on open models. Set up Groq.
• Voyage AI (embeddings) - a large monthly free allowance, enough to index most libraries. Set up Voyage.
• Qwen - includes a free quota to start. Set up Qwen.
• Together AI - usually gives some free trial credit. Set up Together.

Free models on OpenRouter (one small top-up)

OpenRouter offers a rotating set of free models - their names end in :free, and the line-up is 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 (the minimum) is enough, and the free models then keep working without drawing it down. That single charge also raises your daily limits and leaves a balance you can spend on paid models whenever you need more power.

So one $5 top-up buys ongoing access to a changing menu of free models, plus the option of stronger paid models on demand.

Free and private: run models locally

If your computer is reasonably capable, local models (LM Studio or Ollama) are completely free and fully private - nothing leaves your device. They work for both chat and embeddings.

A simple free setup

1. Use Gemini (free tier) as your chat model - or Groq for speed.
2. Use Voyage (free allowance) as your embedding model.
3. Optionally add a one-time $5 to OpenRouter to unlock its rotating free models.

That combination can cover most research use at no ongoing cost.