Skip to content

LLM gateways compared

Every gateway we hold a catalog for, side by side: how many models it carries, how many it publishes prices for, and when we last checked. Counts come from each gateway's own /models endpoint.

Availability data only in this snapshot. This build was generated from FreeRouter's model-map snapshot, which records which gateway serves which model but carries no pricing, context windows, or descriptions. Those columns fill in once the catalog refresh runs against the gateway APIs — see about the data.
Gateway Status Models Text / multimodal With pricing Max context Last checked
OpenRouter The reference catalog for unified LLM access. Live 430
Vercel AI Gateway Vercel's gateway, wired into the AI SDK. Live 373
Cloudflare AI Gateway Edge gateway in front of Workers AI and third-party providers. Live 2,721
OrcaRouter Unified routing with an automatic model picker. Live 195
Requesty Routing policies, caching, and guardrails. Live 706
Concentrate Unified gateway with a bare-id catalog. Live 200
LLM Gateway Open-source, self-hostable, AGPLv3. Live 284
Baseten Dedicated inference for open-weight models. Live 16
Darkbloom Compact OpenAI-compatible gateway. Live 7

Models counts every id in the gateway's catalog, including image, video, and embedding models. Text / multimodal narrows that to models that take a prompt and return text. With pricing is how many of them the gateway publishes a token price for — the rest are served but unpriced in its public catalog.