Frequently asked.
Why is it so much cheaper than list price?
We buy capacity in bulk from the upstream provider and pass most of the discount back. Bulk purchase agreements aren't generally available to individual developers, so aggregating usage lets everyone reach volume pricing.
Is this the real Claude / GPT / Gemini?
Yes. Requests are proxied to the actual upstream API endpoint of each provider. Responses are unmodified. We do not fine-tune, quantize, or route to third-party clones.
What happens if my request fails?
Failed requests are not billed. If the upstream returns an error, you get the same error transparently — no wrapping, no silent retries.
Do you store my prompts?
By default we log only request metadata (model, tokens, latency, status) for billing and abuse prevention. We do not persist request or response bodies unless you explicitly enable request logging for debugging. See the privacy policy for details.
Do you support streaming?
Yes. Both SSE streaming and standard responses work exactly as the upstream provider supports them.
Do you support function calling / tool use / vision?
Yes, wherever the upstream model supports it. We proxy the request unchanged, so if the model supports a feature, so do we.
How do I refund unused credit?
Within 14 days of top-up, contact support@llmrelay.dev with your account email. We refund unused balance to the original payment method. See the refund policy for the full terms.
What models are available?
All major models from Anthropic, OpenAI, Google, Meta and DeepSeek. Full list on the pricing page and via GET /v1/models.
Is there a rate limit?
Yes, per-account rate limits apply to prevent abuse. Default limits are generous enough for most workloads; contact support if you need higher throughput.
Can I get an invoice for business use?
Yes. Every top-up automatically generates an invoice available in your dashboard.
Not answered here? Reach out at support@llmrelay.dev.