Use Cherry Studio to Chat With Several Models

Keep your desktop chats in one place

Cherry Studio is a desktop chat client that lets you work with several AI models from one application. Instead of opening a different website for each service, you can keep your conversations in a familiar interface and choose a model for the task at hand.

The client and the service answering your requests have separate roles. Cherry Studio handles the conversation interface, while an API provider processes your messages and bills for usage. This walkthrough uses MyRelay, a personal, pay-as-you-go relay that gives you one key and one budget for several model families.

Prepare your account and a small test budget

Create a MyRelay account, obtain your API key, and check your available balance before connecting Cherry Studio. Start with an amount you are comfortable spending on a few experiments rather than treating the first session as an unlimited trial.

Keep the key private. It authorizes requests against your balance, so do not include it in screenshots, shared notes, or conversations with a model. If you accidentally expose it, replace it through your provider's account controls.

Download Cherry Studio from its official distribution source and install the version for your operating system. Menu names can change between releases, so use the instructions below as a guide to the relevant settings rather than an exact map of every button.

Add MyRelay as an OpenAI-compatible provider

Open Cherry Studio's settings and find the section for model providers. Add a provider using its OpenAI-compatible option, and give it a recognizable name such as MyRelay. You are selecting a request format here, not limiting yourself to models from a single company.

Enter your MyRelay API key in the provider's key field. Use https://api.myapirelay.com as the starting API base. Check how your Cherry Studio release constructs the final request address: some compatible-provider forms handle the version suffix themselves, while others expect it in the address. Follow the current MyRelay connection guidance and the field's instructions rather than adding suffixes repeatedly.

Save the provider and enable it if Cherry Studio shows a separate activation control. If a connection test is available, run it. A successful connection is useful confirmation, but the next step is still to verify an actual chat request.

Add two models you will actually use

If Cherry Studio can retrieve the provider's model list, refresh it and select only a small starting set. If manual entry is required, copy the exact API model identifiers from MyRelay's current model menu or documentation. A readable product name is not necessarily the identifier accepted by the API.

You could begin with Gemini 2.5 Flash and Claude Haiku 4.5 for everyday questions, short explanations, and small coding discussions. Before choosing between them, check their current prices and availability. Do not assume that a familiar name guarantees the lowest cost for every request.

There is little benefit in filling your sidebar with models you have not tested. Two working choices make it easier to learn the interface and understand your spending.

Send a short prompt and verify the charge

Start a new conversation, select your configured provider and one model, and send a brief request. For example, ask for an explanation of a programming concept in three sentences. Avoid uploading a project or pasting a long document during this initial check.

After the reply arrives, open your MyRelay usage view. Members can inspect each request's tokens and cost. Confirm that the request appears and review the charge before continuing. This connects what you see in the chat window with what you actually spend.

Compare answers without multiplying unnecessary work

To compare models, start a fresh conversation with the second model and use the same short prompt. Separate conversations help keep the comparison from being affected by earlier replies. Judge correctness, clarity, and usefulness alongside price, rather than rewarding the longest answer.

For regular work, keep one default model and switch deliberately. Ask for concise answers when appropriate, avoid repeatedly attaching unchanged material, and start fresh when old conversation history is no longer relevant. Longer histories can increase input usage when they are sent again.

Review your request costs after the first few sessions. With a working connection and a small set of tested models, Cherry Studio becomes a practical desktop workspace rather than another reason to lose track of your AI budget.