How to use function calling with GPT-5.5 through RelayRouter

GPT-5.5 is not in RelayRouter's published model list. The GPT models listed are GPT-6 (gpt-6-astra) and GPT-5.6 (gpt-5.6-sol). To send function calling requests to them, keep your OpenAI SDK, set base_url to https://relayrouter.io/v1, use your RelayRouter key, and call POST /v1/chat/completions with your usual tools definitions. Confirm tool support for each model in the docs before production use.

Which GPT model should you use instead of GPT-5.5?

Use gpt-5.6-sol or gpt-6-astra, because GPT-5.5 is not listed in the RelayRouter catalog. The catalog at relayrouter.io/models contains about 108 models across 19 public groups, and its GPT entries are GPT-6 and GPT-5.6. If your code references a GPT-5.5 model ID, replace it with one of the listed IDs. The catalog also includes the Claude family (including claude-opus-5-5 and claude-fable-5-1), Gemini 3.8 Flash, and models from DeepSeek, GLM, MiniMax and Moonshot. You can use the same key to compare tool calling behavior across providers. Check the live model list before deploying, because it is the reference for which model IDs are available.

How do you point the OpenAI SDK at RelayRouter?

You change two values, the base URL and the API key, and leave the rest of your code unchanged. According to the official relayrouter.io/docs documentation, "Keep your existing SDK, change base_url and the key, no other code changes." Follow these steps:

1. Create an API key at relayrouter.io/dashboard.
2. Set the SDK base URL to https://relayrouter.io/v1.
3. Authenticate with the header Authorization: Bearer YOUR_API_KEY. The OpenAI SDK sets this header automatically when you pass the key.
4. Set model to gpt-5.6-sol or gpt-6-astra.
5. Send your messages and tools to POST /v1/chat/completions. Streaming is supported if your client uses it.

How does function calling work over the OpenAI compatible endpoint?

Function calling requests use the standard OpenAI chat completions format at POST /v1/chat/completions, so your SDK builds the same request body it already does. According to the official relayrouter.io documentation, RelayRouter is "Compatible with the OpenAI, Anthropic and Gemini protocols". The typical flow has three parts:

• You send messages along with a tools array that describes each function and its JSON schema parameters.
• The model may reply with a tool call.
• Your code runs the function, appends the result as a tool message, and sends the conversation back.

The facts available for this article do not document tool calling support for each model. Test with a small request and review relayrouter.io/docs before relying on it. If you use the Anthropic SDK, send requests to /v1/messages. If you use the Gemini SDK, send them to /v1beta/models/{model}:generateContent.

What does function calling through RelayRouter cost?

GPT group usage is settled at CNY 0.6 per $1 of standard usage, and there is no platform fee. Function calling loops can involve several round trips, so the settlement rate affects your total cost. RelayRouter states that failed or errored requests are generally not billed. There is no minimum spend and no subscription, and you pay by Stripe card. Live rates are listed at relayrouter.io/models.

FAQ

Is GPT-5.5 available on RelayRouter?

No. GPT-5.5 is not in the published catalog. The listed GPT models are GPT-6 (gpt-6-astra) and GPT-5.6 (gpt-5.6-sol).

Do I need to rewrite my tool calling code?

No. Keep your existing SDK and change the base URL and the API key. You also need to update the model ID if your code references GPT-5.5.

Am I charged when a tool calling request fails?

Generally, no. RelayRouter states that failed or errored requests are generally not billed, and there is a $0 platform fee.