Skip to main content

​
Overview

Every model below is available through both Gateway endpoints — Chat Completions and Messages — under the same creator/model-name id. Prices are in USD per 1 million tokens, and you are billed at the requested model’s rates regardless of how the request is served. For the live catalog, query GET /models; the tables below show the current catalog.
Cached input is billed separately from fresh input: cache reads are billed at the discounted per-model rate shown in the “Cache read” column, and cache writes at the rate shown (or at the input rate where no dedicated write rate is listed). Reasoning tokens are billed at the output rate.

​
Anthropic Models

​
OpenAI Models

When a request’s input exceeds 272K tokens, these models switch to long-context rates for both input and output on that request:
OpenAI models support service_tier: "flex", which halves both input and output token rates when the request is served on the flex tier.

​
Google Models

​
xAI Models

When a request’s input exceeds 200K tokens, xai/grok-4.5 switches to long-context rates for both input and output on that request: 4.00input/4.00 input / 12.00 output.

​
Perplexity-Hosted Models

Open-source models hosted by Perplexity — the perplexity/ prefix reflects who serves the model, not who created it. Currently Moonshot AI’s Kimi K3 and Z.AI’s GLM 5.2:

​
Listing Models Programmatically

The response lists every available model sorted by id, including each model’s base token prices — the endpoint is the self-describing source for the catalog and base rates, while long-context thresholds and flex-tier multipliers appear only in the tables above. Requesting a model that is not in the catalog returns a 400 naming the invalid model — the catalog is also the allowlist.

​
Next Steps

Quickstart

Make your first Gateway API call.

Pricing & Billing

How credits, billing, and usage tiers work across the platform.