DeepSeek
fā bù
发布
V4
：
ràng
让
AI
xiàng
像
fā duǎn xìn
发短信
yī yàng
一样
pián yi
便宜
DeepSeek Releases V4: Making AI as Cheap as Sending a Text Message
zuì jìn最近 ，zhōng guó中国de的kē jì科技gōng sī公司 DeepSeekfā bù发布le了zuì xīn最新de的 AIdà mó xíng大模型 V4 。
Recently, Chinese tech company DeepSeek released its latest large AI model, V4.
qù nián去年 ，zhè这jiā家gōng sī公司de的 R1mó xíng模型céng曾ràng让quán shì jiè全世界gǎn dào感到zhèn hàn震撼 ，ér而jīn nián今年de的 V4zé则biǎo xiàn表现dé得gèng jiā更加wěn zhòng稳重hé和chéng shú成熟 。
Last year, the company’s R1 model shocked the world, and this year’s V4 is more steady and mature.
V4mó xíng模型de的yí gè一个zhòng yào重要tè diǎn特点shì是 “jí zhì极致de的xiào lǜ效率 ” 。
One important feature of the V4 model is its “extreme efficiency.”
suī rán虽然tā它de的guī mó规模fēi cháng非常dà大 ，dàn但zài在shí jì实际gōng zuò工作shí时 ，tā它zhǐ只huì会diào yòng调用yī xiǎo bù fèn一小部分hé xīn核心néng lì能力 。
Although it is very large in scale, when it actually works, it only uses a small part of its core capabilities.
zhè zhǒng这种zuò fǎ做法jiù就xiàng像shì是yī jiā一家cōng míng聪明de的gōng sī公司 ，zhǐ只zài在xū yào需要de的shí hòu时候ān pái安排zuì最hé shì合适de的zhuān jiā专家qù去gōng zuò工作 。
This approach is like a smart company that sends the most suitable experts to work only when they are needed.
zhè yàng这样zuò做bù jǐn不仅néng能bǎo zhèng保证xìng néng性能 ，hái还néng能dà fú dù大幅度jiàng dī降低chéng běn成本 。
This not only ensures performance, but also greatly reduces costs.
cǐ wài此外 ， V4de的chǔ lǐ néng lì处理能力yě也biàn dé变得gèng qiáng更强le了 。
In addition, V4’s processing ability has become stronger.
tā它kě yǐ可以yí cì xìng一次性dú wán读完yī zhěng běn一整本shū书huò zhě或者fēi cháng非常zhǎng长de的jì shù技术bào gào报告 。
It can read an entire book or a very long technical report all at once.
yǐ qián以前 ，yòng hù用户xū yào需要bǎ把zhǎng长wén zhāng文章fēn chéng分成hǎo jǐ cì好几次fā gěi发给 AI ，xiàn zài现在yī cì一次jiù就néng能jiě jué解决 。
In the past, users had to split long articles into several parts to send to AI, but now one time is enough.
zhè这ràng让rén men人们zài在chǔ lǐ处理fù zá复杂gōng zuò工作shí时 ，biàn dé变得gèng jiā更加qīng sōng轻松hé和gāo xiào高效 。
This makes it easier and more efficient for people to handle complex tasks.
DeepSeekhái还jì huà计划zài在wèi lái未来shǐ yòng使用gèng更xiān jìn先进de的guó chǎn国产xīn piàn芯片 ，jìn yí bù进一步jiàng dī降低fú wù服务jià gé价格 。
DeepSeek also plans to use more advanced domestic chips in the future to further lower service prices.
mù qián目前 ，téng xùn腾讯hé和ā lǐ阿里děng等dà gōng sī大公司dōu都biǎo xiàn表现chū出le了qiáng liè强烈de的tóu zī投资yì yuàn意愿 。
At present, major companies such as Tencent and Alibaba have shown strong interest in investing.
zhè这shuō míng说明 ，shì chǎng市场fēi cháng非常kàn hǎo看好 DeepSeekzhè zhǒng这种zhuī qiú追求xiào lǜ效率hé和kāi yuán开源de的lù xiàn路线 。
This shows that the market is very optimistic about DeepSeek’s path of pursuing efficiency and open source.
duì对 DeepSeeklái来shuō说 ，jìng zhēng竞争de的mù biāo目标bù不zhǐ shì只是zuò做 “zuì qiáng最强 ”de的mó xíng模型 。
For DeepSeek, the goal of competition is not just to build the “strongest” model.
tā men他们zhēn zhèng真正de的mèng xiǎng梦想 ，shì是ràng让 AIchéng wéi成为xiàng像shuǐ水hé和diàn电yī yàng一样pián yi便宜de的jī chǔ shè shī基础设施 。
Their real dream is to make AI a basic infrastructure as cheap as water and electricity.
tā men他们xī wàng希望zài在wèi lái未来 ，diào yòng调用yī cì一次dǐng jí顶级 AImó xíng模型de的fèi yòng费用 ，néng能xiàng像fā发yī tiáo一条duǎn xìn短信nà yàng那样dī低 。
They hope that in the future, the cost of calling a top AI model once will be as low as sending a text message.
DeepSeekde的 V4gào sù告诉wǒ men我们 ， AIjì shù技术zhèng zài正在cóng从 “zhuī qiú追求shén qí神奇 ”zhuǎn xiàng转向 “zhuī qiú追求shí yòng实用 ” 。
DeepSeek’s V4 tells us that AI technology is shifting from “pursuing magic” to “pursuing practicality.”
ràng让měi gè rén每个人dōu都néng能yòng dé qǐ用得起 、yòng dé hǎo用得好 ，cái才shì是rén gōng zhì néng人工智能fā zhǎn发展de的zhēn zhèng真正fāng xiàng方向 。
Making AI affordable and easy for everyone to use is the true direction of artificial intelligence development.
By Easy Mandarin News|AI-assisted