资源点直接通兑 Token,一个套餐覆盖云资源与 AI 调用Resource Points Direct Redemption for Tokens — One Package Covers Cloud Resources and AI Calls


AI free token/AI免费积分

本文来源:https://mp.weixin.qq.com/s/NY5iXOPcnt0ecmnt1wgGiw
若有版权问题,请来信告知:13951431913#139.com 或在右方提交评论。

 

做个带 AI 的应用,最烦的不是写代码。写代码有热情,但算账没有。

云资源多少钱、AI 模型问一句多少钱……每一项都要单独买、单独管,项目还没上线,账单先看晕了。

这次升级解决了什么

CloudBase 对计费方式做了一次升级:资源点现在可以直接抵扣 AI 模型的 Token 用量。

资源点成为统一的计费单位,云资源消耗与 AI 调用都从同一个池子扣减。你只需要管理一份额度。

图片
资源点通兑:一个池子云资源和 AI Token 都能付

资源点是什么

一句话:资源点 = 云开发的统一货币。数据库、云函数、存储、流量、AI 调用所有云资源,可以通兑,如果不够,还可以加购资源点包或开按量付费。

扣费顺序:套餐 → 资源包 → 按量。

1000 点 =  1 元,10000 点 = 10 元

新用户:选择一个套餐即可开始

新用户进入控制台后,选择任一全栈开发 Plan 套餐,资源点计费默认开启,即可直接开发。

我们提供三档套餐,按规模选择其一即可,无需自行组合:

  • • 个人版 ¥19.9/月(限时优惠) 40,000资源点相当于约 200 万次数据库调用,或约1.9 亿 tokens*(DeepSeek-V4-Flash 输入),附 4GB 容量,自己练手、第一次试用够了。
  •  
  • • 标准版 ¥199/月,330,000资源点
  • 相当于约 1600 万次数据库调用,或约 15亿 tokens*(DeepSeek-V4-Flash 输入),附 300GB 存储流量,能接微信支付。适合独立开发者、一人公司。
  •  
  • • 企业版 ¥999/月 1,500,000资源点相当于约 7500 万次数据库调用,或约69亿 tokens*(DeepSeek-V4-Flash 输入),附独享数据库 + 企业微信集成。适合团队与企业级应用。

*Token换算说明:按 DeepSeek-V4-Flash 直供模型、80% 输入命中上下文缓存估算(缓存命中 20 点/百万 tokens,未命中 1000 点/百万 tokens)。实际缓存命中率因使用场景而异,仅作参考,具体以官网资源点价格文档为准。

三档套餐,选一个就够,不用再搭配

开发完成后,你将获得可公开访问的线上地址,可直接分享给任何人访问,而非仅限本地预览。

老用户:余额与有效期完整保留

我们已为存量用户准备了平滑的迁移方案,原有 Token 资源包持续有效。

  • • 旧套餐用户:在控制台手动切换至资源点计费后,Token 资源包余量将自动转换为资源点,有效期原样保留,额度一分不少。
  • • 资源点套餐用户:现有 Token 资源包可继续正常使用;2026 年 8 月 31 日前未消耗的部分,将自动转入通用资源点包,无需任何手动操作。
老用户迁移零损失:旧余额自动转入,一分不少

切换路径:控制台 → 套餐用量页 → 切换资源点计费,一步即可完成。

Token 怎么用资源点付

调用 AI 模型时,输入和输出会换算成 Token,直接从资源点里扣,不用再买 Token 包。

各模型单价以「资源点 / 百万 tokens」标。支持的主流模型如下:

图片

模型完整列表(包括 GLM-5.1、Kimi-K2.6、MiniMax-M3、Qwen3.5-Plus 等)见 AI 控制台 → 生文模型。

跑在套餐里,等于大模型也打折

资源点本身在套餐里有折扣,所以用套餐内的资源点去调 AI,相当于大模型也打了折。

算个账:

DeepSeek-V4-Flash:这个模型的官方标价是:输入 ¥1/百万 tokens,输出 ¥2/百万 tokens。如果你买的是个人版套餐(¥19.9/月,40,000 点),每 1 个资源点相当于 ¥0.00498,比按量便宜一半。

按这个折扣价去调 DeepSeek-V4-Flash,等效价格就是 输入 ¥0.5/百万 tokens,输出 ¥1/百万 tokens。

一次对话输入 5k + 输出 2k tokens,原价 ¥0.009,套餐内实际只花 ¥0.0045。

相当于,一分钱调用2次。

用什么模型不用提前决定

在 AI 控制台启用所需模型后,调用时通过 model 参数指定即可,扣点路径完全一致。需要切换模型也不影响余额管理。

图片

写在最后:如何使用Token


场景一:在 WorkBuddy 里配置自定义模型

step 1:打开对话任务 → 模型配置 → 添加自定义模型Clipboard_Screenshot_1778558112.png#946px #645pxClipboard_Screenshot_1778558273.png#946px #645pxClipboard_Screenshot_1778558440.png#946px #645px

配置项 描述 值
Base URL 大模型 API 的请求地址 https://.api.tcloudbasegateway.com/v1/ai/cloudbase
API Key 访问大模型 API 的密钥 控制台 → 环境配置 → API Key
Model 具体的模型标识符 deepseek-v4-flash 无论选哪个模型,一定要准确复制

 

step2:获取环境id

控制台 → 查看环境 → 复制环境 idXnip2026-05-12_16-18-23.jpg

step3:获取 Base URL 和 API key

控制台首页:AI开发工具 → AI工具中使用Token → 复制 Base URL 和 API Key

图片

AI栏:控制台→ AI → 生问模型 → 复制 Base URL 和 API Key图片

 

step4:点击保存配置

Xnip2026-05-11_17-35-461#909px #563px.jpg

step5:配置完成,在工作去开始对话,验证模型是否生效Clipboard_Screenshot_1778558656909#922px #568px.png

Cursor、Cline、Claude Code 配置方式相同,Provider 选 OpenAI Compatible,填 Base URL 和 API Key 即可。

场景二:小程序端直接调用

图片

 

流式调用,适合实时对话、长文本生成:

const model = wx.cloud.extend.AI.createModel("cloudbase");const res = await model.streamText({data: {model: "deepseek-v4-flash",messages: [{ role: "user", content: "介绍一下李白" }],  }});forawait (const text of res.textStream) {console.log("文本片段:", text);}

换模型只改 model 字段:

model: "minimax-m2.7"// MiniMaxmodel: "glm-5v-turbo"// GLMmodel: "kimi-k2.6"// Kimi

计费应当为开发服务。这次升级的目标只有一个:让你不必为额度分心,把时间留给真正重要的事——把产品做出来。

如有疑问,欢迎在留言区反馈,或点击原文查阅官方文档。

 

 

Source: https://mp.weixin.qq.com/s/NY5iXOPcnt0ecmnt1wgGiw. If there are any copyright issues, please email us at 13951431913#139.com or submit a comment on the right.

Developing an app with AI is not the most frustrating part — coding is actually exciting. It’s the billing math that kills the vibe.

How much does cloud storage cost? What about API calls? Every service requires separate payments and tracking, and you’re deciphering the bill before your project even launches.

What This Update Solves

CloudBase has revamped its pricing model: Resource Points now directly offset AI token usage.

This creates a unified currency for all cloud resources — whether it’s database reads, serverless functions, or model queries — all deducted from one pool. You only need to manage a single quota.

What Are Resource Points?

Think of them as the universal coin for cloud development. Need more bandwidth? More model prompts? Simply convert your points. If they run out, you can buy more or enable pay-as-you-go.

The deduction order follows a simple hierarchy: Package → Prepaid Bundle → Pay-as-you-go.

Conversion: 1000 points ≈ ¥1.00, 10000 points ≈ ¥10.00.

For New Users: Start with a Single Plan

Upon entering the console, select any Full-Stack Development Plan. Resource Point billing is automatically enabled, allowing immediate development without configuring individual services.

We offer three tiers — Personal, Standard, and Enterprise — each designed for specific scales of usage; no need to mix-and-match.

• Personal Plan: ¥19.9/month (limited time discount), 40,000 points — supports ~2M DB calls or ~190M tokens (DeepSeek-V4-Flash input), plus 4GB storage. Perfect for beginners or side projects.

• Standard Plan: ¥199/month, 330,000 points — ~16M DB calls, ~1.5B tokens, 300GB I/O, includes WeChat Pay integration. Ideal for solo entrepreneurs or small businesses.

• Enterprise Plan: ¥999/month, 1.5M points — ~75M DB calls, ~6.9B tokens, dedicated DB cluster, Enterprise WeChat integration. For teams and high-demand apps.

*Token estimation based on 80% cache hit rate for DeepSeek-V4-Flash; actual values may vary depending on prompt locality and model efficiency.

After development, your app gets a public-facing URL — no local testing restrictions.

For Existing Users: Seamless Migration

We’ve prepared a smooth transition for current subscribers. Your existing Token Bundles remain valid until expiration.

• If you’re on an old package, switch to Resource Points and your remaining tokens will convert 1:1 into points — no loss of value or validity.

• If already on Resource Points, token bundles continue to work; any unused balance after August 31, 2026, will automatically convert to generic Resource Points.

To switch: Go to Console → Plan & Usage → Switch to Resource Point Billing.

How Does AI Charge in Points?

When you call an AI model, the system calculates input/output length in tokens and deducts from your point balance — no separate billing cycle.

Each model has a point-per-million-tokens rate. Supported models include:

图片

图片

Running AI on a Plan Means Discounts

Since Resource Points are often discounted in packages, using them to pay for AI effectively gives you cheaper model access.

Example: DeepSeek-V4-Flash costs ¥1/M tokens (input) and ¥2/M tokens (output) at retail. On the Personal Plan (¥19.9/40k points), each point is ~0.00498 RMB — nearly half the cost per token. A 5k-token conversation that would cost ¥0.009 now costs only ¥0.0045 — literally 1 point for 2 full exchanges.

You Don’t Need to Pick One Model Forever

Enable any model in the AI Console, then specify it via the `model` parameter at runtime. You can swap models without changing your payment structure or quota management.

图片

Practical Use Cases

Scenario 1: Configuring a Custom Model in WorkBuddy

Step 1: Open Dialog Task → Model Configuration → Add Custom Model.

Step 2: Get your Environment ID from Console → View Environment.

图片

Step 4: Save configuration — no need to restart or reload.

Step 5: Begin dialog to verify the model is active.

step 1:open the dialogue task → model configuration → add custom modelClipboard_Screenshot_1778558112.png#946px #645pxClipboard_Screenshot_1778558273.png#946px #645pxClipboard_Screenshot_1778558440.png#946px #645px

Scenario 2: Direct Mini-Program Calling

console → view environment → copy environment IDXnip2026-05-12_16-18-23.jpg

Use streaming API for real-time chat or long-form generation. Switch models by changing the `model` field in the request payload.

Final Thoughts: Billing Should Serve Development

图片

AI column: console → AI → generating models → copy Base URL and API Key图片

图片

Xnip2026-05-11_17-35-461#909px #563px.jpg

step 5: configuration completed, start the dialogue in the work area, verify if the model is activeClipboard_Screenshot_1778558656909#922px #568px.png

The configuration methods for Cursor, Cline, and Claude Code are the same. Select OpenAI Compatible for the Provider and fill in the Base URL and API Key.

Scenario 2: Direct calling from the mini-program end

图片

Streaming call, suitable for real-time dialogue and long text generation:

换模型只改 model 字段:

The billing should be for development services. The goal of this upgrade is simple: to free you from worrying about quotas so you can focus on what truly matters — building the product.

If you have any questions, feel free to leave a comment or click the original post for official documentation.

图片

暂无评论,快来发表第一条评论吧!No comments yet. Be the first!

📮 需求咨询📮 Inquiry