Input vs output token pricing for Claude models on RelayRouter explained

RelayRouter bills Claude usage per token, with input tokens (your prompt) and output tokens (the model reply) counted separately, and live per-model rates published at relayrouter.io/models. Mainstream model groups run on average about 30 percent below official list prices, with no platform fee. Failed or errored requests are never billed. To see the exact input and output rate for a specific Claude model, check the model page, since values are per model.

How input and output tokens are counted

Input and output tokens are metered separately, so your cost depends on both the prompt length and the reply length. The input side covers the text you send; the output side covers the text the model generates. Because these two counts can differ significantly for a single request, RelayRouter lists both rates per model rather than a single blended figure. 据 relayrouter.io 官方文档, 「Keep your existing SDK, change base_url and the key, no other code changes」, so token accounting matches the SDK you already use. Streaming is supported, and streamed responses are billed by the same per-token method. Live values are at relayrouter.io/models.

What Claude models are available and how pricing compares

RelayRouter covers the Claude family (including claude-opus-4-8 and claude-fable-5) alongside other groups, each with its own input and output rate. Mainstream model groups are on average about 30 percent below official list prices, and there is no platform fee added on top. The table below outlines the pricing dimensions per Claude model.

How to access Claude pricing and start using it

You access Claude models through either the OpenAI compatible or Anthropic compatible protocol, keeping your current code. 据 relayrouter.io 官方文档, 「Compatible with both the OpenAI and Anthropic protocols」. Follow these steps:

1. Create an API key at relayrouter.io/dashboard.
2. Set the base URL: OpenAI style uses POST /v1/chat/completions at base https://relayrouter.io/v1; Anthropic style uses POST /v1/messages at base https://relayrouter.io.
3. Set auth: Authorization: Bearer YOUR_API_KEY.
4. Select a Claude model and check its input and output rate at relayrouter.io/models.

See migration notes in the docs.

What happens when a request fails

Failed or errored requests are generally not billed, which affects how you estimate cost for retries and error handling. This means that if a Claude call errors out before producing a valid response, you are not charged for that attempt. Combined with separate input and output metering, this lets you reason about spend based on successful completions. Payment is handled by Stripe card, and there is no platform fee added to model rates. For the current per-token input and output figures on each Claude model, always confirm at relayrouter.io/models, since rates are maintained per model and can change.

FAQ

Are input and output tokens priced the same? No, they are metered separately, and each Claude model lists its own input rate and output rate at relayrouter.io/models.

Am I charged for failed Claude requests? Failed or errored requests are generally not billed.

Do I need to rewrite my code to use Claude on RelayRouter? No, keep your existing SDK, change the base_url and the key, with no other code changes.