LLM API pricing

Order

Models are sorted by three factors, in order of priority:

1. Knowledge cutoff: Models with the most recent training data appear first. Models without a known cutoff date appear last.
2. Context window: Among models with the same knowledge cutoff, those with larger context windows rank higher.
3. Name: Models that tie on both factors are sorted alphabetically.

Prices

Each row shows the cheapest provider for the model, with input and output rates from that same provider. Rates are base-tier, on-demand prices in USD per 1M tokens. "Open weights" marks a model whose creator publishes the weights and lists no hosted price.

Sorting and filtering

The list opens on the newest 20 models; "Show all" reveals the rest, and so does any search, filter or sort. Click a column header to sort by it, again to reverse it, and a third time to restore the default order. The search, the creator chips and "Open weights" (models whose creator publishes the weights) combine: a model has to match all three. A retired model no provider lists any more is not in the list; its own page stays up.

Search

We match the model name, its creator and the providers that serve it. Matching is partial and case-insensitive, and every word has to match, so "llama novita" shows only Novita's servings of Llama.

Transparency and funding

• Affiliates: Affiliate links are marked. We may earn a commission if you click them, but commissions never affect the order.