Grok 4.7 setup on CheapRelay: what will the token bill cost?

01 / Grok 4.7 setup on CheapRelay: the first questions

Q. What is the short version of the Grok 4.7 setup on CheapRelay?

A. Create an account, add prepaid funds, choose an OpenAI-compatible client, and enter the model id grok-4.7. CheapRelay bills by the real tokens your request uses. There is no monthly fee, and your unused balance does not expire.

Q. Does every new account come with usable credit?

A. Yes. Every new CheapRelay account receives $0.09 in signup credit, and that credit works with any model. That is enough to send a small test request and check the response path before committing a larger top-up.

Q. What does Grok 4.7 cost today?

A. On the CheapRelay price list, Grok 4.7 is $0.5056 per 1,000,000 input tokens, $1.52 per 1,000,000 output tokens, and $0.1264 per 1,000,000 cache-read tokens.

Q. Which client should I use for Grok 4.7?

A. Use an OpenAI-compatible client under the /v1 path. That includes Cursor, Cherry Studio, Cline, Roo Code, OpenCode, the OpenAI SDK, and cURL. Point the client at CheapRelay's API base and use the model id grok-4.7.

Q. Do I need a special Grok client?

A. No. Grok belongs to the group of models that work through the OpenAI-compatible route under /v1. The model-specific restrictions on this page apply to Claude and Gemini, not Grok. That keeps the Grok 4.7 setup familiar if you already use an OpenAI-compatible tool.

Q. Why start with the price rather than the feature list?

A. Because a token bill is measurable. CheapRelay's price list gives you input, output, and cache-read figures, so you can calculate a workload before sending it. For Grok 4.7, the three rates are $0.5056, $1.52, and $0.1264 per 1,000,000 tokens respectively.

02 / Grok 4.7 token prices: what does a realistic example cost?

Q. How much would 100,000 input tokens and 20,000 output tokens cost on Grok 4.7?

A. Here is an example using the current price list:

• ■ Input: 100,000 ÷ 1,000,000 × $0.5056 = $0.05056
• ■ Output: 20,000 ÷ 1,000,000 × $1.52 = $0.0304
• ■ Example total: $0.05056 + $0.0304 = $0.08096

Q. What happens if that example also reads 50,000 cached tokens?

A. Cache reads are charged separately. Here is another example:

• ■ Input: 100,000 ÷ 1,000,000 × $0.5056 = $0.05056
• ■ Output: 20,000 ÷ 1,000,000 × $1.52 = $0.0304
• ■ Cache read: 50,000 ÷ 1,000,000 × $0.1264 = $0.00632
• ■ Example total: $0.05056 + $0.0304 + $0.00632 = $0.08728

Q. How does Grok 4.7 compare with Grok 4.6 on the same example?

A. Grok 4.6 is listed at $0.4653 for input, $1.39 for output, and $0.1167 for cache reads. Using the same 100,000 input, 20,000 output, and 50,000 cache-read example:

• ■ Grok 4.6: $0.04653 + $0.0278 + $0.005835 = $0.080165
• ■ Grok 4.7: $0.05056 + $0.0304 + $0.00632 = $0.08728
• ■ Example difference: $0.08728 − $0.080165 = $0.007115

Q. What would a larger Grok 4.7 example cost?

A. For 1,000,000 input tokens and 500,000 output tokens, the example is $0.5056 + $0.76 = $1.2656. This is why the request mix matters: output tokens carry a higher rate than input tokens on the price list.

Q. Is Grok 4.7 the only model worth pricing before I start?

A. No. Compare the rates before choosing. Qwen3.7 Plus lists input at $0.1208, output at $0.3625, and cache reads at $0.0361. For the 100,000 input and 20,000 output example, Qwen3.7 Plus costs $0.01208 + $0.00725 = $0.01933, versus $0.08096 for Grok 4.7. The example difference is $0.08096 − $0.01933 = $0.06163.

Q. What about a middle-priced alternative?

A. GLM-5.3 lists $0.4583 for input, $1.38 for output, and $0.1153 for cache reads. On the same 100,000 input and 20,000 output example, GLM-5.3 costs $0.04583 + $0.0276 = $0.07343. Grok 4.7 costs $0.08096 on that example, so the difference is $0.00753. These are examples, not forecasts of your own traffic.

03 / Why does CheapRelay publish prices and official-list comparisons?

Q. What is CheapRelay's pricing approach?

A. Price first, then the arithmetic. CheapRelay uses prepaid balances and charges per real token. There is no monthly fee, and the balance does not expire. The price list is where you check the input, output, and cache-read rates before running a workload.

Q. Where is the published gap to official list price?

A. The price list includes official list figures for the Claude and GPT models shown there. For example, Claude Sonnet 5 output is $4.88 on CheapRelay against an official list price of $10. The arithmetic is $10 − $4.88 = $5.12, which is 51.2% below the list figure, matching the 51% shown on the price list.

Q. Does Grok 4.7 have an official list figure in that comparison?

A. The price list does not show an official list price for Grok 4.7, so this guide does not invent one. Grok 4.7 is compared directly with the other Grok and non-Claude, non-GPT models whose prices are published on the list. That keeps every comparison anchored to a visible number.

Q. How does token billing compare with a subscription?

A. A subscription charges you for access over a period regardless of how many tokens you use. CheapRelay charges for the tokens actually used. If your workload is small or irregular, that pay-as-you-go structure can leave less unused spend than paying for a fixed subscription.

Q. Is there a refund if I top up too much?

A. Yes, within the stated window. Unused balance can be refunded within 12 days of the top-up, minus what was spent and the payment fee. That makes a first-day test budget more practical than loading a large balance without checking the setup.

Q. Can referral rewards reduce the cost?

A. The referral program pays 20% of what a directly invited user pays, plus 6% from the second level, into a separate withdrawable rewards balance. The smallest withdrawal is ¥20, paid through Alipay. Referral rewards are not the same as prepaid API credit, so keep the two balances separate in your own records.

04 / Which client paths work, and which do not?

Q. What is the correct client path for Grok 4.7?

A. Use any OpenAI-compatible client under /v1. Suitable choices include Cursor, Cherry Studio, Cline, Roo Code, OpenCode, the OpenAI SDK, and cURL. The model id to enter is grok-4.7.

Q. Can I use Claude models in the same OpenAI-compatible setup?

A. No. Claude models work only with the official Claude Code and Claude Desktop clients, version 2.1.285 or newer. Do not use Cursor, Cherry Studio, Cline, Roo Code, OpenCode, an SDK, or cURL for Claude, and do not use an OpenAI-compatible configuration. Keeping the client path separate prevents a confusing setup failure.

Q. What about Gemini models?

A. Gemini uses its native protocol only, under /v1beta, through Gemini CLI. It is not an OpenAI-compatible route. Grok remains on the /v1 OpenAI-compatible path, so you can configure each model according to its own rule.

Q. Which other models can use the OpenAI-compatible route?

A. GPT models, including those reached through Codex, plus DeepSeek, Kimi, GLM, Qwen, Grok, MiniMax, and the other models listed for that route can use an OpenAI-compatible client under /v1. Examples of those clients are Cursor, Cherry Studio, Cline, Roo Code, OpenCode, the OpenAI SDK, and cURL.

Q. What model ids should I keep handy on day one?

A. For a few routes you may try, the price list provides:

• ■ Grok 4.7: grok-4.7
• ■ Grok 4.6: grok-4.6
• ■ Qwen3.7 Plus: qwen3.7-plus
• ■ GLM-5.3: glm-5.3
• ■ MiniMax M3: MiniMax-M3
• ■ DeepSeek V4 Flash: deepseek-v4-flash

Q. Why not install several clients and test everything at once?

A. A smaller first test is easier to audit. Choose Grok 4.7, configure one OpenAI-compatible client under /v1, send a short request, and confirm that the response and token usage look right. Then compare the resulting spend with the worked examples above.

05 / Limits, balance checks, and a sensible first-day routine

Q. What request limits apply?

A. Each key allows 120 calls per minute, and each account can run 8 calls at the same time. These limits are enough for parallel testing without turning a first-day experiment into an uncontrolled burst.

Q. What does HTTP 402 mean?

A. It means the account balance is at or below zero. Add prepaid funds before the next request. Because CheapRelay charges per real token, keeping a small working balance is the practical way to avoid an interrupted run.

Q. What should I do before sending a large batch through Grok 4.7?

A. Here is a short example routine:

• ■ Use the $0.09 signup credit for one small request.
• ■ Confirm the client uses the /v1 OpenAI-compatible path and the model id grok-4.7.
• ■ Price a sample workload with the $0.5056 input, $1.52 output, and $0.1264 cache-read rates.
• ■ Top up only the amount you are comfortable testing.
• ■ Watch the balance, keeping in mind the 120-calls-per-minute and 8-concurrent-call limits.

Q. Is the balance still there if I do not use it right away?

A. Yes. The prepaid balance does not expire. If you change your mind within 12 days of a top-up, unused balance can be refunded after subtracting the amount spent and the payment fee.

Q. What makes this Grok 4.7 setup straightforward?

A. The pieces line up: one familiar OpenAI-compatible path under /v1, a published price for each token type, prepaid billing with no monthly fee, and signup credit for a first test. CheapRelay gives you the numbers before you commit the budget, and the worked examples above let you calculate the expected line items yourself.