About
About GeniusRelay
GeniusRelay is one OpenAI-compatible endpoint in front of 599 models from 37 providers, 43 of them uncensored or abliterated. Every model page carries its own input, output and cached-input price before you send a request.
Why it exists
Relays usually hide the rate until after integration, and uncensored models are buried under mainstream ones. We do the opposite: prices are on the page, capability filters are top-level, and the whole catalogue is available as JSON so a script can evaluate it without a sales call.
How pricing works
Prepaid credits, charged per token actually used: uncached input, cached input and output are distinct line items. There is no subscription and no platform fee. Per-key spend caps and expiry dates bound what an application can consume.
Prices shown on the site are reference rates. When the upstream pricing feed is unavailable, the pages fall back to the stored rate; the charge recorded in the console is the authoritative number.
Where the models come from
Models are served through upstream routes run by the providers and by compute partners. Client requests stay in the OpenAI chat completions format, so the same key reaches text, vision, reasoning, tool-calling and long-context models; the model ID selects the route.
What it does not do
- No free tier. New accounts get a small trial credit, then it is prepaid only.
- No embeddings or image generation endpoint yet — chat completions only.
- No fine-tuning and no weight hosting.
- No SLA on individual upstream routes. Retries and failover are the client's job to configure.
Who it is for
Developers comparing models before committing, teams that need an uncensored tier next to their mainstream models, and anyone who wants price and context visible instead of inferred. If you need a signed contract and a dedicated cluster on day one, this is not that product yet.
Contact
Start in the console, or read the docs first: create an account · documentation · pricing
Elsewhere: github.com
Machine-readable resources: llms.txt and the public model JSON. /llms.txt · /data/models.json