夜雨聆风学习资料网

ARTICLE · 1155345

经济学人精读:AI时代的“算力代币”与大模型集市的崛起

经济学人精读:AI时代的“算力代币”与大模型集市的崛起

一、外刊原文及翻译

原文1:计量单位的演变与Token的定义

A UNIT OF account is a useful thing. Even before money was common, people tried to come up with standardised ways of accounting for things. In “The Iliad” the value of armour fashioned from bronze or gold was expressed as a number of oxen (nine for bronze, 100 for gold). Fur pelts, giant stones and piles of salt have all been used as measuring standards. The universe of artificial intelligence is governed by its own unit: the token. These are snippets of text, typically four characters, that an AI model reads or responds with.

翻译:

计价单位是一项非常实用的发明。甚至在货币普及之前,人们就试图寻找对物品进行标准计量的手段。在《伊利亚特》中,青铜或黄金打造的铠甲价值是以牛的头数来衡量的(青铜甲值9头牛,黄金甲值100头牛)。兽皮、巨石和盐堆都曾被用作度量标准。而人工智能的世界则由其独特的单位来支配——Token(代币/标记)。它们是AI模型在阅读或回复时处理的文本碎片,通常约为四个字符。

这里的:

  • unit of account
    :计价单位、记账单位;
  • fashion from
    :由……制成、打造;
  • snippets of
    :碎片、小段。

原文2:Token计价的混乱与巨大价差

The trouble with tokens, however, is that they do not work quite as neatly as cows or pelts did. For one thing, it is unclear in advance just how many a given task might consume. Model-makers also charge dramatically different prices: a popular model offered by DeepSeek, a Chinese startup, is about one-fourtieth the price per token of a cutting-edge one produced by Anthropic, a top American lab.

翻译:

然而,Token带来的麻烦在于,它的运作远不如当年的牛只或毛皮那样一目了然。一方面,用户无法事先准确预知某项具体任务究竟会消耗多少Token;另一方面,各大模型开发商的定价差异极其悬殊:中国初创企业深度求索(DeepSeek)旗下的一款热门模型,其单个Token的调用价格仅为美国顶级实验室Anthropic尖端模型的四十分之一左右。

  • neatly
    :整齐利落地、清晰地;
  • in advance
    :事先、提前;
  • cutting-edge
    :尖端的、最前沿的。

原文3:企业困境与模型集市的应运而生

For firms using AI, this is a recipe for frustration. A simple task, such as asking a model to process invoices, could rack up a huge bill. To insure against this, some businesses cap how many tokens employees can consume. But that may hinder productivity. This muddle has birthed a new kind of business: the model marketplace.

翻译:

对于应用AI的企业而言,这种情况极易让人大为头疼。诸如让模型处理发票这样简单的任务,都可能累积成一笔巨额账单。为了防范这种风险,一些企业对员工可以消耗的Token数量设置了上限,但这又可能阻碍生产力的提升。这种混乱局面催生了一种全新的商业形态——大模型集市(聚合平台)。

  • a recipe for
    :导致……的原因、极易引发……的因素;
  • rack up
    :累积、迅速增加;
  • cap
    (动词):设定上限、限制;
  • muddle
    :混乱、困局。

原文4:OpenRouter的创立与巨额收购

The best known of these marketplaces is OpenRouter, co-founded in 2023 by Alex Atallah. After ChatGPT was released, Mr Atallah began to wonder what kind of world lay ahead. “My biggest question at the time then was: could this be a winner-takes-all market where OpenAI dominates,” he says. He doubted such an outcome, leading to the conclusion that marketplaces would play a central role in the AI industry.

翻译:

在这些模型集市中,知名度最高的是由亚历克斯·阿塔拉于2023年联合创办的OpenRouter。在ChatGPT发布后,阿塔拉先生开始思考未来世界会是何种图景。“我当时最大的疑问是:这是否会变成一个由OpenAI一家独大的‘赢家通吃’市场?”他说道。他对此深表怀疑,并由此得出结论:模型集市将在AI产业中扮演举足轻重的核心角色。

  • winner-takes-all
    :赢者通吃、胜者独揽;
  • dominate
    :主导、占据支配地位。

原文5:动态路由机制与Stripe的天价并购

OpenRouter lets customers easily switch between models. If Anthropic’s Claude is down, they might use one from OpenAI. Or they might swap when a lab is offering a bargain on tokens. OpenRouter charges a small fee for use routed through its platform. Business has been so good that in August Stripe, a payments giant, agreed to buy the marketplace. It will reportedly pay $7.5bn for a firm that is barely three years old.

翻译:

OpenRouter允许客户在不同模型之间自如切换。如果Anthropic的Claude服务宕机,用户可以切换至OpenAI的模型;或者当某家实验室推出Token折扣优惠时,他们也能灵活更换。OpenRouter对经由其平台路由调用的请求收取少量手续费。其业务发展极为迅猛,以至于支付巨头Stripe在8月份同意收购该平台。据报道,Stripe将斥资75亿美元收购这家成立仅近三年的企业。

  • be down
    :(计算机系统或网络)宕机、瘫痪;
  • swap
    :交换、切换;
  • route through
    :经由……路由/分发。

原文6:云巨头的竞争与“中立性”缺失

OpenRouter is not the only model marketplace. Similar services are operated by the big cloud-computing providers. Amazon offers a range of models via Bedrock, billed through existing cloud-service agreements; Google offers the same through its Vertex AI service. But many customers view such offerings with suspicion: the cloud giants, which build their own models and invest in the frontier labs, are hardly neutral.

翻译:

OpenRouter并非唯一的模型聚合平台。大型云计算服务商也运营着类似业务:亚马逊通过Bedrock提供多种模型选择,费用直接通过客户现有的云服务协议结算;谷歌则通过其Vertex AI服务提供同类功能。但许多客户对这些大厂服务心存疑虑:云计算巨头们既自己研发专属模型,又重金投资各家前沿实验室,自身立场很难保持中立。

  • billed through
    :通过……进行结算/开具账单;
  • view… with suspicion
    :对……持怀疑态度;
  • frontier labs
    :前沿AI研发机构。

原文7:对抗“跑分刷榜”与真金白银的投票

OpenRouter first gained a following among developers as it offered a way to play around with lots of different models and a means for quickly gauging which would perform best. After the release of ChatGPT, many AI labs took to “benchmaxxing”—making models that could pass certain tests, such as the SAT, with flying colours, but struggled with more useful tasks. On OpenRouter’s marketplace customers vote with their dollars, creating a fairer ranking system.

翻译:

OpenRouter最初在开发者群体中声名鹊起,是因为它提供了一种可以自由试用多种不同模型、并能快速评估哪款模型表现最优的途径。在ChatGPT发布后,许多AI实验室开始热衷于“刷榜跑分”(benchmaxxing)——即专门针对某些考试(例如SAT测试)进行过度调优,使其能以优异成绩通关,但在处理更实用的任务时却力不从心。在OpenRouter集市上,客户通过真金白银的调用进行投票,从而构建起一套更为公允的排名体系。

  • gain a following
    :赢得追随者、积聚拥趸;
  • play around with
    :摆弄、随意尝试;
  • with flying colours
    :出色地、成绩优异地;
  • vote with one’s dollars
    :用钱包投票(实际付费支持)。

原文8:成本控制诉求与中国高性价比模型的崛起

More recently companies have turned to OpenRouter as a tool for controlling costs. Tellingly, it is the cheap systems offered by DeepSeek and two other Chinese model-makers, Z.ai and Xiaomi, that dominate token consumption on the platform (though an OpenAI model is also currently doing well).

翻译:

近来,越来越多的企业开始将OpenRouter视作控制成本的得力工具。耐人寻味的是,在平台上的Token消耗总量中,占据主导地位的正是由DeepSeek以及另外两家中国模型厂商智谱AI(Z.ai)与小米所提供的低成本系统(尽管OpenAI的一款模型目前的调用量也表现不俗)。

  • tellingly
    :意味深长地、显而易见地;
  • dominate consumption
    :占据消耗量的主导地位。

原文9:“任务熵值”分级法则

Even so, Mr Atallah argues that many businesses remain unsophisticated in their approach to token use. He recommends to his enterprise clients that they divide up tasks into low, medium and high “entropy”, based on how open-ended the work is. Cheap models are often sufficient for low-entropy tasks, such as deciding which model should review some code, and plenty of medium-entropy tasks as well, such as actually reviewing that code. Frontier models can be reserved for high-entropy tasks, like writing the code from scratch.

翻译:

即便如此,阿塔拉先生指出,许多企业在Token的使用策略上依然显得相当粗放。他建议企业客户根据任务的开放性程度,将其划分为低、中、高三种“熵值”(复杂度)。低成本模型往往足以胜任低熵任务(例如决定由哪款模型来审查某段代码),也能应对大量中熵任务(例如实际去审查该代码);而顶尖前沿模型则应专用于高熵任务,例如从零开始编写代码。

  • unsophisticated
    :简单粗放的、不够成熟精细的;
  • entropy
    :熵(借用物理学概念,指任务的混乱度、不确定性或开放性);
  • from scratch
    :从零开始、白手起家。

原文10:避免资源浪费的解耦建议

“The mistake I see a lot of companies make is they just don’t decompose all of these things. It is all the same to them,” says Mr Atallah. “That’s just a massive waste of money.”

翻译:

“我看到许多公司常犯的错误,就是根本不对这些任务进行拆解细分。对他们而言,所有任务都一视同仁地丢给顶级模型,”阿塔拉先生表示,“这纯粹是在极大地浪费金钱。”

  • decompose
    :分解、拆分;
  • a massive waste of money
    :巨大的资金浪费。

二、重点词汇讲解

1. unit of account

  • 词性:
     名词短语
  • 释义:
     计价单位;记账货币
  • 例句:
     Money serves three fundamental functions: a medium of exchange, a store of value, and a unit of account.
  • 考研用法:
     经济学与金融学核心术语,指用于计算商品、劳务以及资产价值的统一度量衡。

2. cutting-edge

  • 词性:
     形容词
  • 释义:
     尖端的;最前沿的
  • 例句:
     The research facility is equipped with cutting-edge medical technology.
  • 考研用法:
     科技类文章高频词,相当于 state-of-the-art 或 advanced,常用于修饰算法、设备或科研突破。

3. a recipe for

  • 词性:
     习语
  • 释义:
     导致……的原因;极易引发……的祸根
  • 例句:
     Imposing unrealistic deadlines on engineers is a recipe for disaster.
  • 考研用法:
     写作与阅读常考,字面为“……的配方/食谱”,引申指某类行为或局势必然会催生某种消极后果(如 a recipe for trouble/failure)。

4. winner-takes-all

  • 词性:
     形容词短语
  • 释义:
     赢家通吃的;胜者全得的
  • 例句:
     The digital economy often exhibits winner-takes-all dynamics due to powerful network effects.
  • 考研用法:
     产业经济与反垄断题材高频表达,形容具有极强网络效应或技术壁垒的市场形态。

5. route

  • 词性:
     动词、名词
  • 释义:
     (动词)按特定路线发送/分发;路由;(名词)路线
  • 例句:
     The server automatically routes incoming network traffic to the nearest data centre.
  • 考研用法:
     在计算机、物流与通信语境中极为常见,强调根据特定规则对数据、请求或物资进行中继调度。

6. neutral

  • 词性:
     形容词
  • 释义:
     中立的;不偏不倚的
  • 例句:
     The platform claimed to be a neutral arbiter, but its algorithmic bias was evident.
  • 考研用法:
     政论与商业博弈核心词,反义词为 biased(有偏见的)或 partisan(偏袒一方的)。

7. with flying colours

  • 词性:
     习语
  • 释义:
     出色地;成绩优异地;大获全胜地
  • 例句:
     She passed the rigorous medical board examinations with flying colours.
  • 考研用法:
     源自航海凯旋时悬挂彩旗的典故,常与 pass、succeed 连用,形容高分过关或极其圆满地完成测试。

8. vote with one’s dollars / feet

  • 词性:
     习语
  • 释义:
     用钱包/脚投票;用实际行动表明偏好
  • 例句:
     If consumers dislike the subscription model, they will simply vote with their dollars and leave.
  • 考研用法:
     经济与社会评论中极具表现力的习语,强调市场力量和个体用购买行为或离去来对抗不合理供给。

9. tellingly

  • 词性:
     副词
  • 释义:
     意味深长地;显而易见地;说明问题地
  • 例句:
     Tellingly, none of the senior executives attended the product launch event.
  • 考研用法:
     外刊高频论证副词,用于引出具有极强说服力、足以揭示事物本质的客观现象。

10. entropy

  • 词性:
     名词
  • 释义:
     熵(物理学/信息论概念);不确定性;混乱度
  • 例句:
     In information theory, entropy measures the unpredictability of a signal or message.
  • 考研用法:
     原为热力学和信息论专业术语,现代管理学与科技评论中常用其隐喻任务的不确定性、无序度或开放自由度。

11. decompose

  • 词性:
     动词
  • 释义:
     分解;拆分;腐烂
  • 例句:
     Complex engineering projects must be decomposed into manageable sub-tasks.
  • 考研用法:
     自然科学中指物质腐败分解,系统工程与计算机科学中特指将复杂的大系统拆解为独立的子模块。

12. from scratch

  • 词性:
     习语
  • 释义:
     从零开始;白手起家;从头做起
  • 例句:
     Building a software architecture from scratch requires significant upfront capital.
  • 考研用法:
     经典考研与日常写作习语,指在没有任何既有基础、模板或现成组件的情况下开展工作。

三、文章主旨

本文深入探讨了生成式人工智能时代以“Token”为核心计量单位所引发的计费困局,以及由此催生的“大模型聚合集市”(Model Marketplaces)新商业形态。

文章首先梳理了人类历史上计价单位的演变(从《伊利亚特》中的牛只到毛皮与盐堆),并指出AI时代的法定计量标准正是Token。然而,Token的使用极具不确定性:单项任务的消耗量难以事先精确估算,且不同模型间的定价差异悬殊(如深度求索DeepSeek与顶级模型相比价差高达数十倍)。这种不可预测性不仅给企业带来了巨大的成本焦虑,也险些抑制员工的实际生产力。

正是在这一计费混乱的背景下,以OpenRouter为代表的第三方模型集市迅速异军突起:

  1. 中立性与灵活路由
    :与自研模型、利益深度绑定的云计算巨头(如亚马逊Bedrock、谷歌Vertex AI)不同,独立的集市平台具备真正的中立性,允许开发者和企业在服务宕机或出现价格优势时,动态无缝切换模型;支付巨头Stripe以75亿美元对其展开的收购,更印证了该聚合赛道的惊人商业价值;
  2. 重塑评价体系与反“跑分刷榜”
    :面对实验室针对标准化测试进行“benchmaxxing”(过度刷榜)的虚假繁荣,集市凭借开发者的实际付费调用量构建出公允的实用排名;
  3. 降本增效的解耦哲学
    :数据显示,高性价比的中国模型(如DeepSeek、智谱Z.ai、小米)正占据绝大部分调用消耗量。创始人阿塔拉更进一步提出“熵值分级”方法论,敦促企业将低熵(确定性高)、中熵与高熵(完全原创)任务严格解耦拆分,把昂贵的前沿模型保留在刀刃上,摒弃不加区分一律调用顶配模型的粗放浪费行为。

文章总结指出,大模型行业的发展正从单纯的技术军备竞赛,逐步过渡到对精细化工程架构、多模型动态调度和成本控制的理性经营阶段。


来源:经济学人

日期:2026年10月10日

原文标题:Artificial intelligence | Token sums

中文标题:人工智能:算力Token时代的计价混乱与模型集市崛起

仅作个人英语学习使用,不构成任何投资建议、技术选型指引或其他专业建议,请勿用于其他商业用途。

相关学习资料