LLM APIs & inference

Most builder tutorials, SDKs, and examples assume this API. It is the first platform to wire up unless you already need a specific other model.

Best for: Shipping a feature on OpenAI models.

If the product needs Claude specifically, use Anthropic's API rather than hoping a chat subscription covers production traffic.

Best for: Apps and agents that need Claude models.

For a builder who wants to switch models without rewriting the client, OpenRouter is the practical third API after the two big labs.

Best for: Trying many models behind one client.

The free tier is large enough to prototype, which makes it the Google API to try before you commit a card.

Best for: Prototyping on Gemini without a large upfront bill.

When the model you want is one Groq hosts and latency matters, Groq is the fast, simple API.

Best for: Low-latency responses from an open model.

Use Mistral when you want that lab's models, an EU provider, or weights you can run yourself. The API and Le Chat are different products.

Best for: Teams that want Mistral's models or a European API.