How to estimate monthly LLM costs on RelayRouter with token math

To estimate monthly LLM costs on RelayRouter, multiply your expected monthly input and output tokens (in millions) by each model's per-1M-token rate. For group-settled models, convert your standard USD usage with the group settlement rate: CNY 0.6 per $1 for GPT and CNY 2.0 per $1 for Claude. Do not add a platform fee, because RelayRouter charges a $0 platform fee with no minimum spend.

How the per-token formula works

Monthly cost equals input tokens divided by 1,000,000 times the input rate, plus output tokens divided by 1,000,000 times the output rate. Directly priced models list their rates per 1M tokens on relayrouter.io/models. For example, deepseek-v4-flash is priced at CNY 1.1 per 1M input tokens and CNY 4.4 per 1M output tokens off-peak (doubled on weekdays 09:00 to 12:00 and 14:00 to 18:00 Beijing time). A workload of 100M input tokens and 20M output tokens per month costs (100 × 1.1) + (20 × 2.2), which is CNY 110 + CNY 44, or CNY 154. Track input and output separately, because output tokens on this model cost twice as much as input tokens. To build the monthly figure, estimate average input and output tokens per request, multiply by daily requests, then multiply by the number of days in the month.

How to estimate Claude and GPT group costs

For the Claude and GPT groups, calculate your usage in standard US dollars and multiply it by the group settlement rate. According to relayrouter.io/models, the GPT group settles at CNY 0.6 per $1 of standard usage and the Claude group settles at CNY 2.0 per $1. The market reference is CNY 6.8 per $1. Standard usage is the dollar value of your tokens at standard rates. Check the live per-model rates before you calculate. The table below applies these rates to $100 of monthly standard usage.

How to collect token counts without changing your code

You can measure real token usage by sending your current SDK traffic through RelayRouter and logging the token counts for each request. According to the official relayrouter.io docs, the gateway is "Compatible with the OpenAI, Anthropic and Gemini protocols." It exposes three endpoints: POST /v1/chat/completions, POST /v1/messages and POST /v1beta/models/{model}:generateContent. The migration guidance from the official relayrouter.io/docs page reads: "Keep your existing SDK, change base_url and the key, no other code changes." Authenticate with Authorization: Bearer YOUR_API_KEY, using a key created in the dashboard. Run representative traffic for a few days, log input and output tokens, and scale the totals to a full month. The catalog covers about 108 models across 19 public groups, so you can compare candidate models on the same traffic.

FAQ about RelayRouter cost estimates

The three questions below cover fees, failed requests and model coverage, which are the factors that commonly change a monthly estimate.

Are there platform fees or subscriptions? No. RelayRouter lists a $0 platform fee, no minimum spend and no subscription, so your estimate is token cost only. Payment is by Stripe card.

Do failed requests count toward the bill? Failed or errored requests are generally not billed. Budget for the requests that complete successfully.

Which models can I estimate? You can estimate the Claude family (including claude-opus-5-5 and claude-fable-5-1), GPT-6 and GPT-5.6, and Gemini 3.8 Flash. DeepSeek, GLM, MiniMax and Moonshot models are also available. Live rates are listed at relayrouter.io/models.