$1 Cherry Studio Budget, Several Models, One Key

01 / Make $1 your practice budget

Give yourself a $1 spending budget for learning Cherry Studio. That is a personal allowance, not a quoted deposit requirement or a promise about how many messages you will get. Your first goal is modest: connect one API key, send a short request, and check what it costs before doing more.

CheapRelay bills for actual input, output, cache reads, and cache writes, with no monthly fee or subscription. Its dashboard itemizes calls and their prices, making it useful for this small experiment. Sign up for CheapRelay, then create an API key in your account. Keep the key private. ([cheapapirelay.com](https://cheapapirelay.com/))

02 / Add CheapRelay as one provider in Cherry Studio

Install the desktop version of Cherry Studio and open Settings, then Model Provider. Cherry Studio keeps connection details under providers; you add models beneath the connection you configure. Its documented setup includes entering credentials, synchronizing models, and enabling the provider. ([cherryai.com](https://www.cherryai.com/docs/en/pre-basic/providers/?utm_source=openai))

• ■ Add a custom provider and give it the name CheapRelay.
• ■ Select the OpenAI-compatible connection type.
• ■ Enter your CheapRelay API key in the API Key field.
• ■ For API Host, enter the HTTPS address for api.cheapapirelay.com, without a path.

Cherry Studio documents that its standard host handling appends /v1/chat/completions to the root address. Avoid entering that complete request path into the host field, which can produce an incorrect address. For this walkthrough, use the OpenAI-compatible connection throughout. ([cherryai.com](https://cherryai.com/docs/en/pre-basic/settings/providers/))

03 / Put two models behind the same key

Click Sync models and add Qwen3.7 Plus and DeepSeek V4 Flash if they appear in your account's available list. Both are listed in CheapRelay's public catalogue. Adding these two gives you a manageable choice for your first comparison. ([cheapapirelay.com](https://cheapapirelay.com/))

If you need to add a model manually, copy its exact model ID from the relay catalogue into Cherry Studio's Model ID field. A display name is for readability; use the supplied ID for the request. After adding models, enable the provider and run Model Check. Retrieved models must actually be added before they appear in the chat selector. ([cherryai.com](https://cherryai.com/docs/en/pre-basic/settings/providers/))

Open a fresh chat, select your first model, and send: Rewrite this sentence in plain English, using no more than 40 words: Our team will undertake a comprehensive evaluation of the proposed changes.

Save the result. Open another fresh chat, choose the second model under CheapRelay, and repeat the same prompt. Judge whether each answer follows the instruction and reads naturally. Keep the prompt identical so your comparison answers a useful question.

04 / Read a $0.00023 example before sending more

At the public rates checked for this article, Qwen3.7 Plus costs $0.1208 per million input tokens and $0.3625 per million output tokens. An illustrative request using 1,000 input tokens and 300 output tokens therefore costs $0.00022955, or about $0.00023, before any separately billed cache activity. Prices can change; check the catalogue before budgeting. ([cheapapirelay.com](https://cheapapirelay.com/))

Those token counts are an example, not a measurement of the test prompt. Find your actual call in the dashboard and record its input tokens, output tokens, cache charges, and total. That receipt tells you more about your workload than a fixed estimate of messages per dollar.

05 / Use a ten-cent checkpoint for the next session

Set a manual review checkpoint at $0.10 of spending. This is your stopping rule, not an automatic account limit. At that point, decide whether the results justify continuing toward your $1 allowance.

• ■ Ask for short answers when a short answer solves the task.
• ■ Use fresh chats for separate comparison tests.
• ■ Check the receipt after a longer prompt or answer.
• ■ Choose a default model after a few useful comparisons.

If the first request fails, check the key, host address, exact model ID, provider switch, and account balance before retrying. Once the connection works, keep the next task small: one email rewrite, one explanation, or one paragraph edit. Expand your usage when the receipts show what your own work costs.