Claude Code pricing: what $10 buys on CheapRelay this month

01 / What does $10 buy for Claude Code pricing on CheapRelay?

Here is the honest way to evaluate Claude Code pricing: pick a budget, pick a model, and divide. CheapRelay publishes per-token prices, so the arithmetic is visible before you commit to anything. There is no monthly fee. You add money to a prepaid balance, and usage is charged against that balance by the token.

For this example, take Claude Sonnet 5.5. Its price is $0.8903 per 1 million input tokens, $4.45 per 1 million output tokens, and $0.0778 per 1 million cache-read tokens. A $10 balance buys:

• ■ 10 divided by 0.8903, which is about 11.23 million input tokens.
• ■ 10 divided by 4.45, which is about 2.25 million output tokens.
• ■ 10 divided by 0.0778, which is about 128.53 million cache-read tokens.

The official list price for Sonnet 5.5 output is $10 per 1 million tokens, so $10 buys 1 million output tokens at that list rate. On CheapRelay, the same $10 buys about 2.25 million. The price list shows official Claude and GPT list prices at 7.2 yuan per US dollar.

That is the point of the price-first approach. CheapRelay is not asking you to accept a vague package. It shows the unit prices, lets you run the division yourself, and keeps the balance for later because it does not expire. Every new account also receives a $0.09 sign-up credit, usable on any model.

02 / What does a month of coding on Sonnet 5.5 cost?

Let us build one example month. Assume you use Claude Sonnet 5.5 for 30 days and consume 10 million input tokens, 2 million output tokens, and 10 million cache-read tokens.

On CheapRelay, the cost is:

• ■ Input: 10 million multiplied by $0.8903 per million = $8.903.
• ■ Output: 2 million multiplied by $4.45 per million = $8.90.
• ■ Cache reads: 10 million multiplied by $0.0778 per million = $0.778.
• ■ Total: $8.903 + $8.90 + $0.778 = $18.581.

Now run the same workload at the official list prices for Sonnet 5.5: input is $2 per million, output is $10 per million, and cache reads are $0.10 per million. That example comes to $20 for input, $20 for output, and $1 for cache reads, or $41.

The difference is $41 minus $18.581 = $22.419 for this example month. That gap comes from the published prices rather than a hidden adjustment. The price list also records Sonnet 5.5 output as 56% under the official list price.

Because billing is per real token, a lighter month costs less and a heavier month costs more. There is no monthly fee to pay regardless. If you stop for a week, the unused balance stays there.

03 / What does the same budget look like on GPT models?

GPT models are also priced per 1 million tokens, so the same $10 test works. GPT-6.1 Sol is $0.1486 for input, $0.7375 for output, and $0.0347 for cache reads. At those rates, $10 buys about 67.29 million input tokens, 13.56 million output tokens, or 288.18 million cache-read tokens.

The official list price for GPT-6.1 Sol output is $10 per 1 million, against $0.7375 on the CheapRelay price list. The published gap is 93% under list. For input, the official list is $2 per million, compared with $0.1486 here. Cache reads are $0.10 at list and $0.0347 here.

GPT-6 Astra is listed at $0.5639 input, $2.82 output, and $0.0569 cache read. Its official list prices are $10, $50, and $1 respectively, with output shown as 94% under list. GPT-5.6 Sol is $0.3056 input, $1.71 output, and $0.0264 cache read, against official list prices of $4, $20, and $0.40, with output shown as 91% under list.

For GPT models, including Codex, use an OpenAI-compatible client under the /v1 path. Examples include Cursor, Cherry Studio, Cline, Roo Code, OpenCode, the OpenAI SDK, and cURL. The model ids to enter are gpt-6.1-sol, gpt-6-astra, and gpt-5.6-sol.

04 / Which model should you choose, and which client fits?

Start with the work, then compare the two numbers that matter: input and output. For example, Qwen3.7 Plus charges $0.1208 input and $0.3625 output. MiMo V2.6 Pro charges $0.1736 input and $0.3472 output. Qwen3.7 Plus has the lower input figure, while MiMo V2.6 Pro has the lower output figure. Neither has an official list price on this price list, so the comparison stays between the two models.

The same approach works elsewhere. DeepSeek V4 Flash is $0.2389 input and $0.9556 output. Hunyuan 4 is $0.2861 input and $0.9139 output. DeepSeek V4 Flash costs less per input token, and Hunyuan 4 costs less per output token. GLM-5.3 is $0.4583 input and $1.38 output. Grok 4.6 is $0.4653 input and $1.39 output. Kimi K3 is $0.9333 input and $4.67 output. Gemini 3.1 Pro is $1.12 input and $6.49 output.

Client choice is simple when you follow the protocol:

• ■ Claude models use only the official Claude Code and Claude Desktop clients, version 2.1.285 or newer. They do not use Cursor, Cherry Studio, Cline, Roo Code, OpenCode, an SDK, cURL, or an OpenAI-compatible setup.
• ■ Gemini models use the native protocol only, under /v1beta, through Gemini CLI. Gemini 2.5 Flash is $0.1778 input, $1.42 output, and $0.0181 cache read. Gemini 3.8 Flash is $0.95 input, $4.28 output, and $0.0944 cache read.
• ■ GPT, DeepSeek, Kimi, GLM, Qwen, Grok, MiniMax, and the other listed models use an OpenAI-compatible client under /v1.

CheapRelay publishes model ids such as qwen3.7-plus, MiMo-V2.6-Pro, deepseek-v4-flash, hy4-preview, glm-5.3, grok-4.6, kimi-k3, gemini-2.5-flash, and gemini-3.8-flash, so you can match a name to a price without guesswork.

05 / How do signup credit, limits, refunds, and referrals work?

Here are the remaining operational answers in short form:

• ■ What does a new account start with? A $0.09 sign-up credit, usable on any model.
• ■ Is there a monthly fee? No. Billing is prepaid and pay per real token, and the balance does not expire.
• ■ What are the call limits? 120 calls per minute per key and 8 calls at once per account.
• ■ What does HTTP 402 mean? The balance is at or below zero, so add funds before the next request.
• ■ Can I get unused money back? Unused balance can be refunded within 12 days of the top-up, minus what was spent and the payment fee.
• ■ What do referrals earn? 20% of what a directly invited user pays, plus 6% from the second level, paid into a withdrawable rewards balance. The smallest withdrawal is ¥20, via Alipay.
• ■ Where do requests go? The API base is https://api.cheapapirelay.com, and the site is https://cheapapirelay.com.

Put the pieces together and the buying decision becomes arithmetic rather than sales language. Choose a model, read its per-token price, multiply by your expected token count, and compare the result with the official list price where one exists. CheapRelay publishes that information plainly, charges by the token, and keeps the setup straightforward: prepaid balance, no monthly fee, published limits, and a clear client-to-model mapping.