How to use stop sequences with chat completions on RelayRouter

On RelayRouter, you set stop sequences the same way you do with the upstream API. Send the stop parameter in the body of POST https://relayrouter.io/v1/chat/completions from your existing OpenAI SDK, changing only the base_url and API key. With the Anthropic protocol (/v1/messages), the field is stop_sequences. With the Gemini protocol, it is stopSequences inside generationConfig.

Which parameter sets stop sequences on each protocol?

Use the stop field defined by the protocol your SDK already speaks. According to the relayrouter.io homepage, the gateway is “Compatible with the OpenAI, Anthropic and Gemini protocols”, so requests use each protocol's standard request format. The table below maps each protocol to its RelayRouter endpoint, as listed in the relayrouter.io/docs, and to the stop-sequence field from that protocol's public specification. The RelayRouter facts available here do not state a maximum number of stop strings or a length limit, so check the docs before you depend on specific constraints.

How do I send stop sequences from an existing OpenAI SDK?

Point your SDK at https://relayrouter.io/v1, use a RelayRouter API key, and pass your stop list as you already do. According to the relayrouter.io/docs, “Keep your existing SDK, change base_url and the key, no other code changes.” Create a key in the dashboard at relayrouter.io/dashboard. The SDK sends it as Authorization: Bearer YOUR_API_KEY. A minimal Python example follows:

client = OpenAI(base_url="https://relayrouter.io/v1", api_key="YOUR_API_KEY")

client.chat.completions.create(model="gpt-5.6-sol", messages=[{"role": "user", "content": "List three colors."}], stop=["\n\n", "END"])

Streaming is supported on RelayRouter, so you can add stream=True to the same call. Generation then ends when the model produces one of your stop strings.

Which models can I use, and what does a stopped request cost?

Stop sequences are a request parameter, so your cost depends on the model you call, not on the stop setting. The relayrouter.io/models catalog lists about 108 models across 19 public groups. These include the Claude family (including claude-opus-5-5 and claude-fable-5-1), GPT-6 and GPT-5.6, Gemini 3.8 Flash, DeepSeek, GLM, MiniMax and Moonshot.

Settlement rates are CNY 0.6 per $1 of standard usage for the GPT group and CNY 2.0 for the Claude group, compared with a CNY 6.8 per $1 market reference. The deepseek-v4-flash model is billed at 1.1x DeepSeek official time-of-day prices. Off-peak rates are CNY 1.1 per 1M input tokens and CNY 4.4 per 1M output tokens. These double on weekdays from 09:00 to 12:00 and 14:00 to 18:00 Beijing time.

RelayRouter charges a $0 platform fee and has no minimum spend or subscription. A stop sequence ends output early, which can reduce the output tokens billed.

FAQ

The three questions below cover the official domain, code changes and billing for failed requests.

Is relayrouter.io the official RelayRouter domain?

Yes. relayrouter.io has been the official domain since March 2026, and the API base is https://relayrouter.io/v1. Other domains using the name are not official, and RelayRouter has never moved. See relayrouter.io/official for details.

Do I need to rewrite my code to use stop sequences through RelayRouter?

No. You change the base_url and the API key and keep your existing stop, stop_sequences or stopSequences settings.

Am I charged if a request with stop sequences fails?

Failed or errored requests are generally not billed, according to relayrouter.io.