
When the “Netflix of Apps” Starts Measuring AI in Credits
In early August 2026, MacPaw made two major moves.
The first was a strategic partnership with Liquid AI to bring edge-side AI inference and local memory capabilities into its AI assistant Eney. Powered by Liquid AI’s Elix inference system and local memory architecture, Eney will be able to operate offline. Once the end-to-end inference stack matures, the technology will be opened to developers on Setapp, while also integrating cloud-based models from providers such as Google to become a unified AI application gateway.
The second move was even more significant:
Within Setapp, the subscription-based Mac application marketplace with more than 150,000 paying users, MacPaw began testing a credit-based AI usage model.
Users consume credits based on their remaining balance and the complexity of AI tasks they perform.
At first glance, these look like two separate product updates.
But viewed together, the direction becomes much clearer.
Setapp is evolving from:
“One subscription, hundreds of Mac apps included.”
into:
A next-generation AI application distribution platform with its own pricing and billing infrastructure.
This is not merely a feature update.
It signals a deeper transition:
The subscription economics of AI agent marketplaces are reaching a turning point.
Credit Systems: Turning “Unlimited Usage” Into Metered AI Consumption
The traditional Setapp model was simple:
Pay $9.99 per month, and access hundreds of applications without worrying about individual pricing.
It became known as:
“Netflix for Mac applications.”
Unlimited access. Unlimited usage.
But AI applications operate under completely different economics.
A single request to GPT-5.6 Luna and a request to Claude Opus 4.8 can have a token cost difference of more than 50 times.
If AI platforms continue using an unlimited subscription model, one of two things happens:
The platform absorbs unsustainable costs.
Developers stop integrating expensive AI capabilities.
MacPaw’s solution is a credit-based system.
The AI Enthusiast plan costs $22.99 per month and includes 6,000 credits.
The AI Expert plan costs $35.99 per month and includes 9,000 credits.
Existing users receive between 3,750 and 7,500 credits depending on their plan.
Credit consumption depends on:
Message length
Task complexity
Model capability
Computing resources required
Simple tasks consume fewer credits.
Complex AI workflows consume more.
This is not simply selling credits.
It is creating a pricing layer for AI intelligence.
Every AI operation becomes measurable.
Every AI capability becomes assignable a value.
The deeper change is about who controls pricing power.
Traditional app stores allow each application developer to set prices independently.
But Setapp’s credit experiment moves toward:
“A unified AI consumption currency controlled by the platform.”
MacPaw is bringing distribution power and billing power into the same ecosystem.
Token Markets: Another Form of AI Credit
From a broader perspective, Setapp’s credit system and Unexhub’s token marketplace are solving the same fundamental challenge:
How do we measure, price, and distribute AI capability?
Setapp credits represent the consumer-side abstraction.
Users purchase credits.
Users spend credits.
The system hides the complexity underneath:
Which model was used?
How many tokens were consumed?
What was the actual infrastructure cost?
Users do not need to know whether Claude Opus 4.8 costs $5 per million tokens or $0.50 per million tokens with caching.
They only need to know:
“How many credits does this task require?”
Unexhub’s token marketplace represents the supply-side abstraction.
It functions as a token aggregation and compute trading infrastructure layer, allowing API prices to dynamically adjust according to supply and demand.
Its core mechanisms include:
Instant Settlement
AI suppliers receive revenue based on actual token consumption, reducing cash-flow pressure.
Zero-Commission Incentives
Early suppliers can join with lower platform costs, encouraging supply-side participation and improving market liquidity.
Dynamic API Bidding
Token prices fluctuate according to real-time demand.
During low-demand periods:
Developers can access cheaper AI capacity.
During peak periods:
Suppliers can capture higher-value demand.
Setapp answers:
“How much should users pay for AI capability?”
Unexhub answers:
“How should suppliers price AI computing power?”
One is demand-side abstraction.
The other is supply-side pricing infrastructure.
Together, they form the foundation of a future AI economy.
The Turning Point for AI Agent Subscription Pricing
Setapp’s credit experiment reveals a major trend:
AI subscriptions are moving from fixed pricing toward usage-based economics.
Over the past two years, AI products have experimented with many pricing methods:
Monthly subscriptions
User seats
Credit packages
Token billing
But the industry has lacked a universal measurement system.
A single “AI conversation” in one application may cost ten times more than another application, while users remain unaware of the difference.
Setapp’s credit system attempts to create a unified measurement layer.
Regardless of which model runs underneath, users interact with one simple concept:
Credits.
This provides two advantages:
For users:
Easier budgeting
Clearer value perception
Less technical complexity
For platforms:
Greater flexibility in choosing models
Easier cost optimization
Ability to route tasks dynamically
As long as one credit represents a consistent level of AI capability, platforms can switch between different model providers without changing the user experience.
This is where token marketplaces become critical.
Unexhub’s token trading infrastructure provides the supply-side foundation behind this transformation.
When tokens can be priced dynamically like electricity, AI application platforms can purchase computing power more efficiently and convert it into user-friendly credit systems.
The future of AI subscription pricing is not about cheaper models.
It is about:
Better infrastructure for measuring, pricing, and distributing intelligence.
Engineer Perspective: Two Abstraction Layers, One Direction
Setapp credits and Unexhub’s token marketplace are essentially two sides of the same transformation.
User Layer: Credits
Credits make AI:
Understandable
Budgetable
Comparable
Users do not need to understand token economics.
They only need to know:
“How many credits do I have left?”
and:
“Can these credits complete my task?”
Infrastructure Layer: Tokens
Token markets make AI computing:
Tradable
Price-discoverable
Liquid
Suppliers do not need to build their own billing infrastructure.
They simply provide token capacity to the market.
Between these two layers lies the most valuable power:
Distribution and pricing control.
MacPaw aims to control AI application distribution and billing through Setapp.
Because whoever controls AI application distribution controls the next generation of app marketplaces.
Unexhub aims to control another critical layer:
Token pricing and circulation.
Because whoever controls token markets controls the infrastructure economy of AI.
The relationship is simple:
Credits are tokens from the user’s perspective.
Tokens are credits from the market’s perspective.
When these two layers mature and connect, AI agent marketplaces will no longer depend on individual model providers setting prices.
Instead, prices will emerge from market supply and demand.
That is the real turning point.
The old story:
One subscription, hundreds of apps.
The new story:
One subscription, unlimited access to AI capability — priced dynamically by the market.
Unexhub: Building the Token Marketplace for the AI Economy
Unexhub is building a global token aggregation and distribution compute marketplace.
The platform connects AI suppliers across different scales, including major AI providers such as:
Gemini
ChatGPT
Claude
DeepSeek
GLM
as well as high-quality AI models from individual developers and small-to-medium businesses.
Through:
Supplier competition
Intelligent routing
Market-based pricing
Real-time settlement
Unexhub enables enterprises and developers to access AI capabilities at significantly lower costs.
On Unexhub:
Developers pay based on actual token consumption;
AI resources compete through market pricing;
Users can access AI services at discounts as low as approximately 30% of market prices;
Suppliers can integrate once and receive real-time revenue sharing from API usage.
The future question is no longer:
“Can you afford the strongest AI model?”
The real question is:
“Can you access AI intelligence in the most efficient way?”
当“应用界的 Netflix”开始用积分计量 AI
2026 年 8 月初,MacPaw 做出了两项重要动作。
第一项,是与 Liquid AI 达成战略合作,为旗下 AI 助手 Eney 引入端侧推理系统 Elix 和本地记忆系统。
这意味着 Eney 将具备更强的离线运行能力,在无需持续依赖云端的情况下完成部分 AI 推理任务。
随着这套端到端 AI 推理体系逐渐成熟,相关技术栈未来还将开放给 Setapp 平台上的开发者使用,同时连接 Google 等云端模型服务,打造一个统一入口式的 AI 应用生态。
第二项动作,则更加值得关注:
在拥有超过 15 万付费用户 的订阅制 Mac 应用商店 Setapp 中,MacPaw 开始测试一种全新的 AI 计费方式:
基于积分(Credit)的 AI 使用模式。
用户根据自己的积分余额,以及任务复杂程度,消耗对应数量的积分完成 AI 操作。
如果把这两件事放在一起看,会发现 Setapp 正在发生根本变化。
过去的 Setapp 是:
“一个订阅,上百款 Mac 应用随便使用。”
它被称为:
“Mac 应用界的 Netflix。”
但未来的 Setapp,可能正在转变为:
一个拥有 AI 分发能力、计费能力和资源调度能力的新型 AI 应用平台。
这并不是一次普通的产品更新。
它释放出的信号是:
AI 智能体应用市场的订阅模式,正在进入新的阶段。
积分制:从“无限畅饮”走向“按量消费 AI 能力”
过去 Setapp 的商业模式非常简单:
用户每月支付固定订阅费用,即可访问平台中的大量 Mac 应用。
这就是典型的:
“Netflix 式订阅”。
用户不关心每一次使用成本。
平台负责承担资源消耗。
但 AI 应用完全不同。
因为 AI 背后存在真实的计算成本。
一次 GPT-5.6 Luna 调用,与一次 Claude Opus 4.8 调用之间,Token 成本可能相差几十倍。
如果 AI 平台继续采用传统“无限使用”模式:
结果只有两个:
第一:
平台承担越来越高的 AI 调用成本。
第二:
开发者因为成本压力,不敢继续加入高级 AI 功能。
因此,MacPaw 选择了积分体系。
目前:
AI 爱好者方案
月费:22.99 美元
包含:6000 积分
AI 专家方案
月费:35.99 美元
包含:9000 积分
老用户则根据套餐不同,每月获得约:
3750~7500 积分。
积分消耗取决于:
消息长度
任务复杂程度
调用模型能力
所需计算资源
简单任务:
消耗更少积分。
复杂 Agent 工作流:
消耗更多积分。
这并不是简单地“卖积分”。
本质上,它是在建立:
AI 能力的统一计价层。
过去:
用户购买的是应用。
未来:
用户购买的是智能。
每一次 AI 操作,都可以被衡量。
每一种 AI 能力,都可以被定价。
真正变化的是:谁掌握 AI 定价权?
传统应用市场中:
每个开发者拥有自己的定价体系。
应用价格由应用开发者决定。
但 Setapp 的积分模式正在改变这一结构。
未来可能变成:
平台统一定义 AI 能力价格。
用户购买积分。
平台负责调度不同 AI 服务。
不同模型之间的成本差异,被隐藏在积分体系之后。
用户不需要知道:
Claude Opus 4.8 输入价格是多少。
也不需要知道:
缓存命中后 Token 成本是多少。
用户只需要知道:
“完成这个任务,需要消耗多少积分。”
这意味着:
MacPaw 正在同时掌握:
AI 应用分发权
AI 能力计费权
而这两项能力,将成为 AI 时代应用市场竞争的核心。
词元交易市场:另一种形式的“积分体系”
如果从更大的产业视角来看:
Setapp 的积分体系,与 Unexhub 正在构建的 Token 交易市场,本质上解决的是同一个问题:
如何衡量、定价和流通 AI 能力。
两者只是处于不同环节。
Setapp:消费端的 AI 抽象层
Setapp 积分解决的是用户侧问题。
用户:
购买积分。
消耗积分。
获得 AI 服务。
整个过程屏蔽了底层复杂性:
使用了哪个模型?
消耗多少 Token?
实际计算成本是多少?
用户看到的是:
“完成任务需要多少积分。”
而不是:
“后台调用了多少 Token。”
Unexhub:供给端的 Token 定价层
Unexhub 构建的是另一侧基础设施:
一个 Token 聚合与分发的算力交易中心。
它解决的是:
AI 能力如何被供应。
核心机制包括:
即时结算
帮助 AI 供应商根据 Token 使用量快速获得收益。
降低传统 API 销售中的资金等待周期。
零佣金激励
吸引更多 AI 供应商加入生态。
让更多模型资源进入市场。
提升整体流动性。
API 动态竞价
让 Token 价格根据市场需求实时变化。
低峰时期:
开发者获得更优惠价格。
高峰时期:
供应商获得更合理收益。
简单来说:
Setapp 解决:
用户应该如何购买 AI 能力。
Unexhub 解决:
AI 能力应该如何被市场定价。
一个连接需求端。
一个连接供给端。
两者共同构成未来 AI 经济的重要基础设施。
AI 智能体订阅模式正在迎来转折点
Setapp 的积分实验,揭示了一个重要趋势:
AI 应用订阅正在从:
固定价格模式
转向:
动态计量模式。
过去两年:
AI 产品出现了大量不同收费方式:
月订阅
企业席位
Token 计费
积分套餐
但行业一直缺少一个统一价值尺度。
同样一次“AI 对话”:
在不同应用中,
背后的 Token 消耗可能相差数倍甚至十倍。
但用户无法感知。
积分体系试图解决这个问题。
无论底层使用:
Claude、
Gemini、
GPT、
DeepSeek,
用户最终面对的是统一的:
积分消耗。
这样:
用户更容易理解成本。
平台也拥有更大的调度空间。
平台可以根据任务:
自动选择更合适的模型。
选择更低成本的算力。
选择更优供应渠道。
工程师视角:两个抽象层,同一个方向
从技术角度看:
Setapp 积分体系和 Unexhub Token 市场,是 AI 经济中的两个抽象层。
用户层:积分
积分让 AI 变得:
可理解
可预算
可比较
用户不需要学习 Token 经济。
只需要知道:
“我的积分还剩多少。”
“这些积分够不够完成任务。”
基础设施层:Token
Token 市场让 AI 算力变得:
可交易
可定价
可流通
供应商无需自己建立复杂计费体系。
只需要提供 Token 能力。
真正关键的是:
两者之间的控制权。
谁控制:
AI 应用分发。
谁控制:
AI 能力计费。
谁就掌握下一代 AI 应用市场。
MacPaw 希望通过 Setapp 掌握:
AI 应用入口。
而 Unexhub 希望通过 Token 市场掌握:
AI 算力流通入口。
积分是用户看到的 Token,Token 是市场看到的积分
未来 AI 经济可能形成这样的关系:
用户看到的是积分。
市场看到的是 Token。
积分负责简化消费体验。
Token 负责实现市场定价。
当这两个体系逐渐连接:
AI 应用价格将不再由单一模型厂商决定。
而会由:
市场供需关系决定。
这才是真正的变化。
过去的故事:
一个订阅,上百款应用随便用。
未来的新故事:
一个订阅,按需购买 AI 能力,而 AI 能力价格由市场发现。
Unexhub:构建 AI 时代的 Token 交易基础设施
Unexhub 正在打造全球 Token 聚合与分发的算力交易中心。
平台已经连接不同规模的 AI 供应商,包括:
Gemini
ChatGPT
Claude
DeepSeek
GLM
等大型 AI 模型服务商。
同时,也支持个人开发者、中小企业提供的优质 AI 模型资源。
通过:
AI 供应商竞价
智能路由
市场化定价
实时结算
帮助企业用户和开发者降低 AI 使用成本。
在 Unexhub:
开发者按照实际 Token 使用量付费;
多供应商竞争推动更优价格;
用户可享受最低约市场价格 3 折的调用优惠;
AI 供应商可一键接入渠道,并获得实时调用分成。
未来,企业竞争的关键不再只是:
“谁拥有最强模型。”
而是:
谁能够以最低成本,将 AI 能力转化为真正生产力。