> ## Documentation Index > Fetch the complete documentation index at: https://docs.sinarcloud.com/llms.txt > Use this file to discover all available pages before exploring further. # AI subscription plans > Pay a fixed weekly price for a set number of DeepSeek V4.1 Flash tokens. How plans count tokens, what happens when they run out, and a worked usage example. An AI subscription plan gives you a set number of tokens for one model every week, for a fixed price paid from your balance. Your requests use the plan's tokens first. Once they're used up, requests are charged to your balance at the model's normal prices. This page explains the plans, how they count tokens, and how a week of usage plays out. ## The plans Every plan on sale today is weekly and covers DeepSeek V4.1 Flash, model alias `deepseek/deepseek-v4-1-flash`. Plans come in two families, shown as tabs on the **Subscriptions** page: * **AI Plans** count every token of a request, including the input the model reads from its cache. * **Cache-Free AI Plans** don't count tokens read from the cache. They count only new input and output, and give fewer tokens for the same price. These are the plans on sale on 1 November 2026: | AI Plans | Tokens a week | Unused tokens | | - | - | - | | DeepSeek Flash 100M | 100M | Expire at the end of the week | | DeepSeek Flash 300M | 300M | Expire at the end of the week | | DeepSeek Flash 500M | 500M | Expire at the end of the week | | DeepSeek Flash 1B | 1B | Expire at the end of the week | | DeepSeek Flash 2.5B No Expiry | 2.5B | Never expire | | DeepSeek Flash 5B No Expiry | 5B | Never expire | | DeepSeek Flash 10B No Expiry | 10B | Never expire | | Cache-Free AI Plans | Tokens a week | Unused tokens | | - | - | - | | DeepSeek Flash 2M Cache-Free | 2M | Expire at the end of the week | | DeepSeek Flash 6M Cache-Free | 6M | Expire at the end of the week | | DeepSeek Flash 10M Cache-Free | 10M | Expire at the end of the week | | DeepSeek Flash 20M Cache-Free | 20M | Expire at the end of the week | | DeepSeek Flash 50M Cache-Free No Expiry | 50M | Never expire | | DeepSeek Flash 100M Cache-Free No Expiry | 100M | Never expire | | DeepSeek Flash 200M Cache-Free No Expiry | 200M | Never expire | Plans can change. The **Subscriptions** page always shows the plans on sale to you, with their current prices in your account's billing currency. ![The Subscriptions page with the AI Plans tab selected next to Cache-Free AI Plans, each with 7 plans. Seven plan cards from DeepSeek Flash 100M to DeepSeek Flash 10B No Expiry each show a weekly price, the tokens of DeepSeek V4.1 Flash every week, a list of what the plan includes and a Subscribe button. The No Expiry plans add Unused tokens never expire and Used it all? Renew early and a new period starts at once.](https://mintcdn.com/sinar-cloud/mho1oQe0vttS15ny/images/ai-subscription-plans.png?fit=max\&auto=format\&n=mho1oQe0vttS15ny\&q=85\&s=7bcc9b41618794b87f4bed8e49ac38cf) The AI Plans tab of the Subscriptions page, with prices in US dollars ![The Subscriptions page with the Cache-Free AI Plans tab selected. Seven plan cards from DeepSeek Flash 2M Cache-Free to DeepSeek Flash 200M Cache-Free No Expiry each show a weekly price, the tokens of DeepSeek V4.1 Flash every week, Cached tokens don't use the allowance, and a Subscribe button.](https://mintcdn.com/sinar-cloud/mho1oQe0vttS15ny/images/ai-subscription-plans-cache-free.png?fit=max\&auto=format\&n=mho1oQe0vttS15ny\&q=85\&s=7491d8d137337b9a9d900bea3ac35349) The Cache-Free AI Plans tab, with prices in US dollars ### No Expiry plans Plans whose names end in **No Expiry** keep their unused tokens. Tokens left at the end of a week carry over to the next week and never expire, even after the subscription ends. If you use up a week's tokens early, you can renew right away instead of waiting for the week to end. See [Renew early](#renew-early). On the other plans, tokens left at the end of the week expire, and the next week starts with a fresh allowance. ## How plans count tokens Plan tokens are token credits for DeepSeek V4.1 Flash. Each week's allowance appears on the [Token credits](https://docs.sinarcloud.com/ai/token-credits.md) page with the source **Plan allowance**. A request takes tokens from the plan like this: | Part of the request | AI Plans | Cache-Free AI Plans | | - | - | - | | New input | Counts | Counts | | Cached input, read from the cache | Counts | Doesn't count | | Cache writes | Counts | Counts | | Output, reasoning included | Counts | Counts | On AI Plans, a request takes its total tokens. On Cache-Free AI Plans, it takes its total tokens minus the cached input. For what each kind of token is, see [How tokens are counted](https://docs.sinarcloud.com/ai/tokens.md). When the plan has enough tokens left, it pays for the whole request, including the cached input and the Rp 0.6 minimum charge. Nothing is charged to your balance, on either family. When the plan has fewer tokens left than the request takes, the plan pays its share of the request and your balance pays the rest at the model's normal prices. The plan's share is the tokens it has left divided by the tokens the request takes. See [When the plan runs out](#when-the-plan-runs-out). Good to know: * **Which requests use the plan:** every request your account sends to `deepseek/deepseek-v4-1-flash`, from any API key, the [Playground](https://docs.sinarcloud.com/ai/playground.md) and tools such as Claude Code. Other models, including DeepSeek V4.1 Flash's ZDR version, are charged as usual. * **Other token credits:** if you also have token credits for DeepSeek V4.1 Flash, for example from a gift, the tokens that expire soonest are used first. Tokens that never expire are used last, oldest first. * **Promotions:** a promotion on the model discounts only what your balance pays after the plan. A promotion that makes the model free applies first, so the plan's tokens stay untouched while it runs. * **An empty balance:** requests to the model still go through while the plan has tokens left, even when your balance is zero or below. * **No automatic failover:** while you have plan tokens for the model, a request that fails isn't retried another way. You get the error and can send the request again. ## Example: a week of usage This example follows the coding session from [Example 6 on the token counting page](https://docs.sinarcloud.com/ai/tokens.md#example-6-a-coding-session-paid-from-token-credits) on two plans with the same price: **DeepSeek Flash 100M** from AI Plans and **DeepSeek Flash 2M Cache-Free**. The session sends four requests: | Request | New input | Cached input | Output | | - | - | - | - | | 1. First message of a coding session | 20,000 | 0 | 1,500 | | 2. Next turn in the same session | 5,032 | 19,968 | 800 | | 3. A turn deep in a long agent session | 9,300 | 132,700 | 700 | | 4. A short chat question | 300 | 0 | 400 | ### On DeepSeek Flash 100M Every token counts, so each request takes its total tokens: | Request | Taken from plan | Left in plan | Paid from balance | | - | - | - | - | | 1. First message | 21,500 | 99,978,500 | Rp 0 | | 2. Next turn | 25,800 | 99,952,700 | Rp 0 | | 3. Long session turn | 142,700 | 99,810,000 | Rp 0 | | 4. Short question | 700 | 99,809,300 | Rp 0 | ### On DeepSeek Flash 2M Cache-Free Cached input doesn't count, so each request takes its new input and output: | Request | Taken from plan | Left in plan | Paid from balance | | - | - | - | - | | 1. First message | 21,500 | 1,978,500 | Rp 0 | | 2. Next turn | 5,832 | 1,972,668 | Rp 0 | | 3. Long session turn | 10,000 | 1,962,668 | Rp 0 | | 4. Short question | 700 | 1,961,968 | Rp 0 | * The session takes 190,700 tokens from DeepSeek Flash 100M and 38,032 from DeepSeek Flash 2M Cache-Free. * On the Cache-Free plan, the 152,668 cached tokens of requests 2 and 3 cost nothing. They're part of each request's price, and the plan pays the whole price. * Paid from your balance without a plan, the four requests would cost Rp 34.77232 at DeepSeek V4.1 Flash's prices on 1 November 2026. Prices can change, but the tokens a plan takes don't depend on them. ### Choose a plan family At the same price, an AI Plans plan gives 50 times the tokens of the Cache-Free plan. Which one lasts longer depends on how much of your input comes from the cache: * A turn like request 3 takes 142,700 tokens from DeepSeek Flash 100M, which then lasts for 700 such turns a week. It takes 10,000 tokens from DeepSeek Flash 2M Cache-Free, which lasts for 200. * A Cache-Free plan lasts longer only when more than 98% of your tokens are read from the cache. To estimate for your own usage, open [Usage](https://docs.sinarcloud.com/ai/usage.md) and compare your **Tokens processed** with your cached input on the **Tokens** tab. ### When the plan runs out Later that week, DeepSeek Flash 2M Cache-Free has 4,000 tokens left, and the session sends another turn like request 3. The turn takes 10,000 tokens, more than the plan has left: ```text theme={null} tokens the request takes = 9,300 + 700 = 10,000 plan's share = 4,000 ÷ 10,000 = 40% tokens the plan covers = 40% × 142,700 = 57,080 of 142,700 full price = see Example 6 = Rp 12.888 plan pays = 40% × Rp 12.888 = Rp 5.1552 balance pays = Rp 12.888 − Rp 5.1552 = Rp 7.7328 ``` The plan now has 0 tokens left. Until a new week starts, every request to DeepSeek V4.1 Flash is charged to your balance at the model's normal prices, as if you had no plan. The [Usage](https://docs.sinarcloud.com/ai/usage.md) page shows how much of each request the plan covered: open **Show details** on the **Log** tab, and look at **Token credits** under **Charge breakdown**. ### Carry over and renew early On a No Expiry plan, the same week goes differently. Say you're on **DeepSeek Flash 2.5B No Expiry**: 1. **Week 1:** you use 1.8B of your 2.5B tokens. The 0.7B left carry over. 2. **Week 2:** the renewal gives you 2.5B new tokens, so you have 3.2B. The 0.7B carried over are used first, because the oldest tokens that never expire go first. 3. **Midweek:** you've used all 3.2B. Requests are charged to your balance, and the subscription's page shows **This period's allowance is used up** with a **Renew now** button. 4. **Renew now:** you pay the plan's price again. A new week with 2.5B fresh tokens starts at once, and your next renewal moves to a week from then. The days left in the week you were in aren't refunded. If you don't renew early, the plan renews at the end of its week as usual. ## Subscribe to a plan **1. Open Subscriptions** In the console sidebar, under **Account**, click **Subscriptions**. Choose the **AI Plans** or **Cache-Free AI Plans** tab. **2. Choose a plan** Click **Subscribe** on the plan you want. The **Subscribe to** dialog, named after the plan, shows the price, what the plan includes, the amount **Due now**, your **Wallet balance** and the **Balance after**. **3. Top up if your balance is short** If your balance doesn't cover the price, the dialog says how much to top up and shows **Top up wallet**. Pay by QRIS, then click **Back to the plan**. See [Add balance](https://docs.sinarcloud.com/billing/top-up.md). **4. Agree and pay** Read the terms in the dialog, tick **I've read and understood how SinarCloud processes tokens**, and click the **Pay** button with the price, for example **Pay Rp 15,000**. The full price is paid from your balance, the first week starts right away, and its tokens are ready to use. You subscribe in the console; API keys can't start a subscription. You can hold one plan in each family at a time, so an AI Plans plan and a Cache-Free plan together. To move to another plan in the same family, [change your plan](#change-your-plan) instead of subscribing again. ## Manage your subscription Your plans appear under **Your subscriptions** at the top of the **Subscriptions** page, with the tokens left this week. Click a plan to open its page. It shows the tokens left, when unused tokens expire or that they carry over, the current period, the next renewal and what each week includes. ### Renewals A week is 7 days, counted from the moment you subscribed, not a calendar week. Shortly before each week ends, SinarCloud charges the plan's price from your balance, and the new week starts the moment the old one ends. This repeats until you cancel. Your available balance must cover the full price. If you used AI on credit and your balance is below zero, the renewal needs that amount too. A day before a renewal, SinarCloud emails the account's owners if your balance is short, and the subscription's page shows **Top up before your renewal** with a **Top up wallet** button. If your balance doesn't cover the renewal: 1. SinarCloud tries again every hour until the week ends. A QRIS top-up retries it at once. 2. If it's still unpaid when the week ends, the subscription shows **Payment due**, and the **Renewal not paid** notice gives a date to top up by. Until then, the plan gives no new tokens. 3. When you top up in time, the plan restarts as soon as the payment comes in, with a full week from then. The time it was unpaid has no plan tokens. 4. If you don't top up by that date, the subscription ends. Tokens that never expire, from a No Expiry plan, stay usable the whole time, and after the subscription ends. ### Change your plan Click **Change plan** on the subscription's page, or **Upgrade** or **Switch to this plan** on another plan in the same family. The dialog says what the change does before you confirm. * **A bigger plan in the same family** starts now. You pay the price difference for the rest of the current week and get the extra tokens for the rest of the week. From your next renewal, you pay the new plan's price. * **A smaller plan** starts at your next renewal. Nothing is charged now. To keep your current plan instead, open **Change plan** again and choose **Keep** with your plan's name. Upgrading from a plan whose tokens expire to a No Expiry plan doesn't change the tokens you already have: they still expire at the end of the week. Only the extra tokens, and the tokens of later weeks, never expire. You can't change to a plan in the other family. Subscribe to it instead: you can hold one plan in each family. ### Renew early On a No Expiry plan, once you've used every token of the plan, carried-over tokens included, the subscription's page offers **Renew now**. Click it, check the amount, and click **Pay** with the price **and renew**, for example **Pay Rp 187,500 and renew**. A new week with a fresh allowance starts at once, and your next renewal moves to a week from then. The days left in the current week aren't refunded. You can renew early again each time the new allowance runs out. ### Cancel Click **Cancel subscription** on the subscription's page, then confirm. The plan stays active until the end of the week you paid for, then ends. Payments aren't refunded. Until the week ends, you can click **Resume subscription** to keep it. If a renewal is unpaid, cancelling ends the subscription at once. ## Where you see plan payments and tokens * **Billing:** each payment for a plan appears on the **Transactions** tab as **Subscription**. See [Balance and billing](https://docs.sinarcloud.com/billing/overview.md). * **Token credits:** each week's allowance is a grant with the source **Plan allowance**. See [Token credits](https://docs.sinarcloud.com/ai/token-credits.md). * **Usage:** a request the plan paid for shows the tokens it covered under **Token credits** in its **Charge breakdown**. See [Usage and spending](https://docs.sinarcloud.com/ai/usage.md). > [card] [How tokens are counted](https://docs.sinarcloud.com/ai/tokens.md) > > Input, cached, output and reasoning tokens, with worked examples. > [card] [Token credits](https://docs.sinarcloud.com/ai/token-credits.md) > > How prepaid tokens apply and expire. This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.