You have tried one lab's API — how to try the rest

You have an OpenAI or Anthropic key, some working code, and a suspicion that there is more out there. There is, and trying it is cheaper than you think — mostly because of one accident of history.

Everything speaks the same protocol. The industry standardised on one API shape, and nearly every provider implements it. In most codebases, pointing your existing client at a different provider is a changed base URL and a new key — your code does not otherwise change. This is what makes shopping around real rather than theoretical.

What is actually different elsewhere. Three things, honestly: price — downloadable models served by competing hosts routinely undercut the big labs several-fold for work that does not need a frontier model; speed — some hosts serve on hardware that streams tokens several times faster, which changes what an interactive product feels like; and variety — coding specialists, small fast models, transcription models, things no single lab offers.

How to compare without a spreadsheet. Each model page here shows every host serving it in one table — price, served context (which genuinely differs between hosts of the same model), measured speed, and what each host says about training on your prompts. Where a host has not said, the table says unknown rather than assuming the answer you would prefer.

One caution as you spread out. More providers means more parties handling your data, and their policies differ more than their APIs do. The privacy columns exist on every provider page for exactly this moment.

Where next: Providers, compared · What is meant by providers