Lowest-Priced Models Across the Web

The ranking above compares input plus output price together, because a model with a cheap input leg and an expensive output one is not cheap for most work. It is a periodic snapshot; for the current rate of any model, see its page in the model catalog.

Any model whose input and output both price at zero appears in the Free Models section at the top of this page. They carry no usage cost at all, which makes them a good fit for development and personal projects.

Both are in the table above with their current per-million-token rates for input and output, side by side. The gap moves when either provider changes list price, so the table is the answer rather than a figure written into this sentence.

Start from the combined input and output cost for the tier you need, then check the ratio: a model that is cheap on input and expensive on output suits short prompts with short answers, and the reverse suits long context. Both legs are priced separately in the table for exactly this reason.

Qwen, GLM, DeepSeek, Moonshot, MiniMax and Kuaishou models are all in the catalog, and several of them sit in the low-price table above. They tend to price well below the US flagships at comparable capability.

List prices sync from upstream daily. Your discount is set on your account and does not move with them, so a provider price cut reaches you rather than widening our margin.