{"@context":"https://schema.org","@type":"NewsArticle","generatedAt":"2026-07-23T06:40:50.084Z","headline":"谁怕中国模型？--Kimi K3 逼近 SOTA，开源模型成本优势引热议","description":"中国开源模型 Kimi K3 在能力上逼近当前最先进水平，引发行业讨论。其 API 价格为每百万输入 token 3 美元、每百万输出 token 15 美元，低于 Sol 的 5 美元和 30 美元。但推理时代 token 并非同质化商品，Kimi 需更多推理 token 才能达到正确答案，实际智能成本取决于模型体积、推理效率、内存效率、服务效率和 token 效率等多重因素。","url":"https://www.aioga.com/news/cmrtwpzse3c55bihzmczi3sbf/","mainEntityOfPage":"https://www.aioga.com/news/cmrtwpzse3c55bihzmczi3sbf/","datePublished":"2026-07-20T23:46:58.488Z","dateModified":"2026-07-20T23:46:58.488Z","inLanguage":"zh-CN","publisher":{"@type":"NewsMediaOrganization","name":"Aioga","url":"https://www.aioga.com"},"citation":["https://stratechery.com/2026/whos-afraid-of-chinese-models","https://aihot.virxact.com/items/cmrtwpzse3c55bihzmczi3sbf"],"canonicalUrl":"https://www.aioga.com/news/cmrtwpzse3c55bihzmczi3sbf/","directAnswer":{"@type":"Answer","text":"Aioga 编辑摘要：中国开源模型 Kimi K3 在能力上逼近当前最先进水平，引发行业讨论。 Aioga 将其归入「技巧观点」方向，重点关注它对真实使用和行业竞争的影响。","url":"https://www.aioga.com/news/cmrtwpzse3c55bihzmczi3sbf/","dateCreated":"2026-07-20T23:46:58.488Z","author":{"@type":"Organization","@id":"https://www.aioga.com/authors/aioga-editorial/#editorial-team","name":"Aioga Editorial Team","url":"https://www.aioga.com/authors/aioga-editorial/"}},"evidence":[{"@type":"CreativeWork","name":"stratechery.com source article","url":"https://stratechery.com/2026/whos-afraid-of-chinese-models","datePublished":"2026-07-20T23:46:58.488Z","provider":{"@type":"Organization","name":"stratechery.com","url":"https://stratechery.com/2026/whos-afraid-of-chinese-models"}},{"@type":"CreativeWork","name":"AIHot archive record","url":"https://aihot.virxact.com/items/cmrtwpzse3c55bihzmczi3sbf","datePublished":"2026-07-20T23:46:58.488Z","provider":{"@type":"Organization","name":"AIHot","url":"https://aihot.virxact.com/items/cmrtwpzse3c55bihzmczi3sbf"}}],"aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","originalPublisher":{"name":"stratechery.com","url":"https://stratechery.com/2026/whos-afraid-of-chinese-models"},"article":{"id":"cmrtwpzse3c55bihzmczi3sbf","slug":"cmrtwpzse3c55bihzmczi3sbf","url":"https://www.aioga.com/news/cmrtwpzse3c55bihzmczi3sbf/","title":"谁怕中国模型？--Kimi K3 逼近 SOTA，开源模型成本优势引热议","title_en":"谁怕中国模特？","summary":"中国开源模型 Kimi K3 在能力上逼近当前最先进水平，引发行业讨论。其 API 价格为每百万输入 token 3 美元、每百万输出 token 15 美元，低于 Sol 的 5 美元和 30 美元。但推理时代 token 并非同质化商品，Kimi 需更多推理 token 才能达到正确答案，实际智能成本取决于模型体积、推理效率、内存效率、服务效率和 token 效率等多重因素。","source":"Hacker News 热门（buzzing.cc 中文翻译）","sourceUrl":"https://stratechery.com/2026/whos-afraid-of-chinese-models","aiHotUrl":"https://aihot.virxact.com/items/cmrtwpzse3c55bihzmczi3sbf","publishedAt":"2026-07-20T23:46:58.488Z","category":"技巧观点","score":69,"selected":false,"articleBody":["There’s a story I tell about my first day in STRT-431 at Kellogg School of Management, the introductory class that every first-year MBA was required to take; I leafed through the readings and case studies and was dismayed that there weren’t any tech companies on the docket. Me being me, I spoke to the professor after class wondering why, and was told that the goal of the course was not to necessarily learn about specific industries, but rather to uncover broadly applicable universal principles that could be applied to any company in any industry.","I did not, as I usually tell the story, find this very satisfactory: to me the nature of tech, particularly the fact that software and distribution had zero marginal costs (and zero transaction costs), was something fundamentally different; putting in zeroes in formulas tends to wreak havoc! I soon realized, however, that that was my opportunity. The fundamental insight undergirding Aggregation Theory：https://stratechery.com/2015/aggregation-theory/ is that zero marginal costs leads to fundamentally different value chains than people once expected from the Internet: centralization and scale in a world where controlling demand mattered more than distributing supply.","What is fascinating about AI, however, is the extent to which those old universal principles are coming back to the forefront. That was never more apparent than this past weekend, when arguments raged on X about the implications of Kimi K3, another open weights model out of China, approaching the state-of-the-art in terms of capabilities. The long and short of it is this: marginal costs are back in a big way, both in terms of short-term implications of state-of-the-art free models, and in terms of the long-term structure of the industry.","One of the most common misconceptions undergirding discussion of open weights models is that they are cheaper — free, even. After all, you can just download the weights, and skip the time and expense and capabilities necessary to create your own model. That is, of course, true, but the “free” in this case is a reference to the amount you need to spend on research and development; R&D is a fixed expense that is independent of the revenue you generate. If you spend $1 million in R&D, it doesn’t matter if you do $100 thousand in revenue or $100 million; you still spent $1 million on R&D (it does, of course, impact your profitability).","What is related to revenue is COGS — cost of goods sold — and COGS is real for AI in a way it hasn’t been for software for a very long time. Specifically, running inference on a model — whether that model be Kimi or Fable — costs money, and the amount of money an AI provider spends on inference is, at least in most business models, directly correlated to revenue. To reuse the above example, generating $100 million versus $100 thousand in revenue will likely require 1,000x COGS. In concrete terms, if it costs 50 cents to generate the tokens that drive $1 in revenue, then $100 million in revenue will have $50 million in COGS; $100 thousand in revenue will only have $50 thousand in COGS.","The point in terms of open weight models is that they are not free to serve. Kimi K3 costs：https://platform.kimi.ai/docs/pricing/chat-k3 $3 per million input tokens, and $15 per million output tokens; that is cheaper than Sol’s $5 per million input tokens and $30 per million output tokens, but that might not even be the right measurement.","Nvidia CEO Jensen Huang has described what Nvidia is building as “token factories”, and from Nvidia’s perspective that framing makes sense. Nvidia GPUs are model agnostic: they generate tokens, and do so in the fastest and most efficient way possible. That leads to measurements like tokens-per-second, time-to-first-token, tokens-per-watt, token cost, etc., and Huang argues that these metrics will be the basis for decision-making.","This is a framing that definitely made sense during the first paradigm of AI, the ChatGPT era, when tokens were delivered straight to the end user. The second paradigm of AI, however, the reasoning era, confounds this measurement. Reasoning entails an explosion in chain-of-thought tokens, and different models need different amounts of reasoning tokens to arrive at the right answer. Kimi, for example, reportedly uses significantly more tokens than Sol, rendering its price advantage moot. Agents introduce a similar dynamic: some models are more efficient than others in terms of the number of tokens they need to execute agentic workflows.","What this means is that tokens are not a commodity. The defining characteristic of a commodity is that it is fungible: a gallon of oil is a gallon of oil; a ton of copper is a ton of copper; a bushel of wheat is a bushel of wheat. A token from one model, however, is not the same as a token from another model. What is fungible is what is constructed from tokens, which is to say intelligence. In other words, if both Kimi and Sol generated the right answer, then that answer is fungible; the difference in tokens generated to get to that right answer is a contributor to a difference in COGS.","The COGS for intelligence is a function of a few different factors:","The reason this matters is that we are rapidly approaching a state in which intelligence for many economically beneficial tasks is in fact a commodity. Anyone building a basic CRUD app：https://en.wikipedia.org/wiki/Create,_read,_update_and_delete, for example, can likely do so using models from multiple providers. And, in a commodity market, the route to profitability is not through charging higher prices — again, you can (or will soon be able to) make the exact same app using multiple models — but rather through having a superior cost structure.","It’s worth stepping through the mechanics here, because, as I noted a few months ago in Amazon’s Durability：https://stratechery.com/2026/amazons-durability/, the dynamics of commodity markets are not something people in tech are generally familiar with:","The key thing to understand is that the marginal cost of producing the commodity differs by supplier. What this means in practice is that the supplier with the worst cost structure ends up selling the commodity at their marginal cost (if they can produce at all); the profits of everyone else depend on the extent to which their cost structure is better than the marginal supplier.","Let’s assume the price elasticity is such that there is demand for 25 units of the commodity at $20. That means:","This isn’t precisely right: the reason why Supplier C will bear the shortfall is because Suppliers A and B will be able to slightly undercut them in price, which will of course affect demand (which is elastic), but it makes the point. Supplier A has a great business, Supplier B has a good business, and Supplier C is going to go bankrupt.","Bankruptcy risk is where fixed costs come back to the forefront: Supplier C has both fixed costs (like potentially R&D spend) and also may have taken on debt to finance the equipment necessary to produce the commodity. It can’t price its commodity with these costs in mind — remember, the market-clearing price approximates the marginal cost of the highest-cost unit needed to satisfy demand — but those costs can absolutely drive the supplier out of business. And, if that supplier goes out of business, then prices go up, until another supplier decides to enter (or the other suppliers expand).","Let’s bring this back to models. Right now, none of the above analysis applies because demand exceeds supply for frontier models, and supply is limited by a lack of compute. This compute shortage doesn’t just mean that a compute supplier like Nvidia makes very large margins, but also that Nvidia’s customers, like SpaceXAI, can turn around and resell compute at high margins as well to a company like Anthropic. Anthropic, meanwhile, can pay the markup because they can sell tokens with a higher markup still.","It’s not just excess demand that gives Anthropic great margins, however: Anthropic and OpenAI likely have among the lowest costs per unit of frontier-quality intelligence, thanks to model capability, serving scale, and token efficiency. They are serving models at a particular capability level for months before their competitors, and are simultaneously applying the best models to optimizing those costs.","It’s also worth noting that the market is not yet treating intelligence like a commodity: demand is for Anthropic and OpenAI specifically, and much less for models that aren’t as good (thus SpaceXAI and Meta selling capacity to Anthropic); one way to think about the push for optimizing cost is that that is a function of defining jobs-to-be-done by intelligence level, such that intelligence buyers can create a market where intelligence is commoditized. In the long run, however, whoever is on the frontier is the best placed to dominate non-frontier markets as well, which are just the frontier minus n-months, i.e. months in which the frontier model makers have been optimizing their cost of serving.","All of this is to say that I think the reaction to Kimi and Chinese models generally is pretty over-blown, at least from an economic perspective. Right now there is a price umbrella that is downstream of the lack of compute; I highly doubt that Chinese models are cheaper to serve on a marginal cost basis, they just seem cheaper because Anthropic and OpenAI are so supply constrained that they are charging far more than they would if there were sufficient supply to meet the demand for intelligence.","Why, then, do the model makers in particular seem so panicked about Chinese models?","First, I think the frontier labs are anchored in a world where training costs dominated their financial modeling. As long as training consumed more GPUs than inference, it was critical to maximize inference revenue to help fund the next training run, which meant charging very high prices for inference.","Going forward, however, I expect the inference market to grow much faster than training costs (and that includes the assumption that training costs will continue to skyrocket), which means they really can make it up in volume. It wasn’t clear this would be the case as recently as eight months ago, but the agent paradigm unlock is so massive that frontier labs should have more confidence that they can not just survive but thrive with lower prices (once they have sufficient compute).","Second, intelligence isn’t in fact a perfect commodity, in part because applied intelligence makes itself smarter. Specifically, whoever is running inference is also collecting data, and that data goes into making the next iteration of the model better. This is, on one hand, all the more reason for the frontier labs to lower prices and increase usage as more compute comes online; on the other hand, this is why companies like Microsoft：https://x.com/satyanadella/status/2076323181154230284 are increasingly obsessed with helping companies run their own models. That is much more viable if Chinese models are a viable alternative.","Third, the other way that frontier labs can not only differentiate from Chinese models but also from each other is by continuing to integrate up into the customer experience. It’s striking the extent to which Claude Code and Codex are proving to be quite sticky; whichever harness you start working with is likely to be the one you stick with, and that figures to be even more the case with non-technical users. And, in the long run, this imperative to move up the stack does mean that frontier models are absolutely a threat to software providers, including Microsoft. On the flipside, the extent to which software companies who currently own the customer experience have access to competitive models is the extent to which they may be able to resist the encroachment of the frontier labs.","Finally, the ideological angle of Anthropic in particular：https://stratechery.com/2026/anthropics-safety-superpower/ is impossible to ignore. This is a company that believes only it can be entrusted with AI, and the existence of open weights alternatives strikes a fatal blow to that presumption.","Kimi isn’t the only new Chinese model; from Bloomberg：https://www.bloomberg.com/news/articles/2026-07-19/alibaba-s-qwen-unveils-preview-of-flagship-ai-model:","Alibaba Group Holding Ltd. shares rose as much as 5.4% on Monday after the company launched a preview version of its flagship Qwen3.8 Max model, describing it as second only to Anthropic PBC’s Fable 5. The Sunday release came only days after startup Moonshot AI unveiled a powerful new offering that’s roiled markets and triggered concern in the US about China closing the gap on global leaders like Anthropic and OpenAI. Qwen3.8 Max has 2.4 trillion parameters, joining Moonshot’s Kimi K3 in the heavyweight class. With 2.8 trillion parameters, K3 rivals top offerings and Alibaba is setting similarly high expectations.","Developers can now access Qwen3.8 Max through Alibaba’s coding platforms, including Qoder. Alibaba plans to make the model open-weight soon, expanding access beyond the preview release. Interest in these made-in-China artificial intelligence systems and models is so high that Moonshot was forced to pause taking on new subscriptions late on Sunday to manage overwhelming demand.","The fact that Qwen3.8 Max will also have open weights is notable. Alibaba stopped releasing weights for its leading edge models earlier this year, but appears to have reverted that change; I suspect that shift was related to last week’s Xi Jinping speech about AI：http://english.scio.gov.cn/topnews/2026-07/18/content_118605932.html that doubled down on the open weights approach:","We should adhere to the principle of openness and win-win and boost innovation-driven development. As a new engine of world economic growth and an accelerator for the shift of growth drivers, AI is moving from the digital world into the physical world. We should seize this rare, historic opportunity to encourage open source, openness, collaboration and sharing. We should facilitate technological innovation, industrial development and scenario-based application of AI. We should make coordinated advances in the transformation and upgrade of traditional industries, the cultivation and growth of emerging industries and forward-looking planning for future industries, so that all sectors and businesses can benefit from AI.","The strategy for China is obvious: commoditize your complements. Note that Xi explicitly ties openness to AI “moving from the digital world into the physical world”; the physical world is the world dominated by China, and the country’s lead in areas like robotics is going to massively benefit from widely available AI models.","Along the same lines, China does not want the U.S. to gain an asymmetric advantage in AI; to the extent that China can weaken the U.S. frontier labs while strengthening any and all potential U.S. adversaries so much the better, and it can benefit from the innovation that will attach itself to an open ecosystem.","By the same token, don’t expect China to do anything about distillation attacks on the frontier labs. I think it is mistaken to attribute all of the success of Chinese labs to distillation, but it’s just as much of a mistake to pretend like distillation doesn’t give Chinese labs a big advantage. That advantage has really come to bear in the last year as post-training reinforcement learning has become increasingly crucial to model performance. Instead of having to fashion reinforcement learning environments from scratch, Chinese labs can simply use frontier labs models as teachers, allowing for rapid improvement at much lower costs (this is not the only reason why Chinese models are cheaper to develop, but it’s a big one).","What is interesting is that one of the most important use cases for Chinese models in the West is itself distillation. Thinking Machines, for example, which just released an open-weight model, relies on Chinese models：https://www.ft.com/content/ef486929-d2c2-480b-8b00-9cb98bda6acf to solve the cold start problem for reinforcement learning. Dean Meyer and Konstantine Buhler wrote an excellent article on X：https://x.com/deanmeyerrr/status/2077834267086729674?s=46 explaining that distillation means that Western open weight models are fundamentally disadvantaged relative to China:","Distillation does not explain China’s entire open-model lead. Chinese labs have world-class researchers, substantial compute, strong pre-trained models, software-hardware codesign, and rapidly improving post-training capabilities. But distillation compresses the costly final gap between a strong base and a near-frontier system. Even if distillation represents a smaller share of a Chinese model’s total capability, it represents a meaningful share of its advantage over American open models.","New enforcement mechanisms will make large-scale distillation harder, slower, and more expensive for Chinese companies. However, enforcement will not eliminate distillation backed by state actors. Every Western frontier advance therefore creates another teacher for Chinese labs. Western builders must either reproduce those capabilities independently or wait to learn from Chinese models. This gap gives Chinese labs a recurring structural advantage over Western companies.","This is a point that bears repeating: because U.S. open weight model makers must follow the frontier labs’ terms of service, they (1) are worse than Chinese alternatives and (2) end up distilling the distillation, just with a detour through Chinese labs. Wouldn’t it be better if western open weight model makers could go to the source?","To that end, here’s an even more interesting question around distillation: why exactly is it bad? After all, what are large language models but the distillation of all of the knowledge on the open Internet, scraped by the frontier labs and distilled into the models that are themselves being distilled? Who is exactly being wronged here?","This entire Article has been an exercise in defusing overreaction to Kimi K3 specifically and Chinese open weight models generally; however, there is one reason to be concerned, and that is cybersecurity. Consider this story from The Stack：https://www.thestack.technology/hugging-face-hacked-turned-to-chinese-llm-for-help-after-us-models-blocked-blue-team/:","Hugging Face said its production infrastructure was breached by an “autonomous” AI agent system early last week. The platform’s security team were initially stymied in their incident response (IR) by unnamed US LLM frontier model guardrails “which cannot distinguish an incident responder from an attacker,” they said. So Hugging Face’s defenders turned instead to the open-source GLM 5.2 model from China’s Z.ai lab – running it on their own infrastructure to analyse the 17,000+ logs, or footprints, that the attackers left behind.","That’s a striking public admission for the New York-headquartered Hugging Face, which lets users collaborate on models, datasets and applications, and which this summer hit the $100 million ARR mark. In an incident report, the company recommended that defenders “have a capable model you can run on your own infrastructure [our italics] vetted and ready before an incident, both to avoid guardrail lockout and to keep attacker data and credentials from leaving your environment.”","It’s difficult to overstate how wrong-headed the Trump administration’s panicked response to Anthropic’s release of Fable was, particularly since it exacerbated Anthropic’s worst tendencies in terms of assuming only they can be trusted with powerful AI. In a world with only one AI, it might make sense to reserve the most powerful cybersecurity capabilities for the U.S. government and trusted allies; however, that’s not the world we live in.","There are and will be models eminently capable of mounting cybersecurity attacks on existing infrastructure, and those models will be — already are — widely available. The best defense — the only viable defense, in fact — will be to make sure defenders have access to the best models as well. Right now defenders are effectively banned from using Fable or Sol for cybersecurity because of Trump administration directives; that means the best alternative is using models from a country which has been trying to weaken our cyber defenses for years. This is insane!","The better course is clear: first, loosen Fable and Sol restrictions on cybersecurity, and second, ensure that U.S. open weight model makers are on an equal playing field with China. Yes, the frontier labs will kick and scream about this, but the Administration should realize that listening to their histrionics has led the U.S. to a position where U.S. companies are dependent on China for their defenses. Let the frontier labs win by being better; don’t let them define safety or security, or pull up the ladder of humanity’s collective knowledge. China is already hard enough to compete with; letting them carry the standard for openness and innovation is simply giving away our biggest advantage.","Stratechery Plus ：https://stratechery.com/stratechery-plus/","The most popular and most important posts on Stratechery by year.","Explore all free articles on Stratechery .","Stratechery Plus ：/stratechery-plus/","Designed with WordPress.：https://wordpress.com/website-builder/?partner_domain=stratechery.com&utm_source=Automattic&utm_medium=colophon&utm_campaign=Concierge%20Referral&utm_term=stratechery.com Hosted by Pressable.：https://pressable.com/?utm_source=Automattic&utm_medium=rpc&utm_campaign=Concierge%20Referral&utm_term=stratechery.com"],"articleImages":[{"sourceUrl":"https://i0.wp.com/goat.passport.online/assets/goat-podcast-cover-plus.png?w=128&h=128&ssl=1","alt":"","afterParagraph":45,"url":"/media/articles/cmrtwpzse3c55bihzmczi3sbf/f3b6788a54e3b11a.webp"},{"sourceUrl":"https://i0.wp.com/sharpchina.fm/assets/SharpChinaArt.png?w=128&h=128&ssl=1","alt":"","afterParagraph":45,"url":"/media/articles/cmrtwpzse3c55bihzmczi3sbf/3da2341b621e9481.webp"}],"mediaStatus":"ok","articleBodyZh":["我讲过一个故事，是关于我在凯洛格管理学院参加STRT-431的第一天，这是一门每个一年级MBA学生都必须上的入门课程；我翻阅了课程的阅读材料和案例研究，却对课程中没有任何科技公司感到失望。像我这样的人，下课后我去找教授询问原因，他告诉我，这门课程的目标并不是一定要了解特定行业，而是发现可以广泛应用的通用原则，这些原则可以应用于任何行业的任何公司。","我并没有像我通常讲故事那样对此感到非常满意：对我来说，科技的性质，特别是软件和分销的边际成本为零（且交易成本为零）的事实，是根本不同的；在公式中加入零往往会引起混乱！然而，我很快意识到，那正是我的机会。支撑聚合理论（Aggregation Theory：https://stratechery.com/2015/aggregation-theory/）的基本洞察是：零边际成本会导致价值链与人们曾经对互联网预期的完全不同：在需求控制比供应分发更重要的世界中，集中化和规模化。","然而，令人着迷的是，人工智能在多大程度上让这些旧的通用原则重新回到前台。上周末，这一点比任何时候都更为明显，当时关于Kimi K3——另一个来自中国的开源权重模型——在能力上接近最先进水平的影响，X上争论激烈。简而言之，情况如下：边际成本已经强势回归，不论是从最先进免费模型的短期影响来看，还是从行业的长期结构来看。","关于开放权重模型讨论中最常见的误解之一是，它们更便宜——甚至是免费的。毕竟，你可以直接下载权重，而跳过创建你自己模型所需的时间、费用和能力。当然，这是真的，但这里的“免费”是指你在研发上需要花费的金额；研发是一项固定开支，与您产生的收入无关。如果你在研发上花了100万美元，无论你产生10万美元还是1亿美元的收入，都无所谓；你仍然在研发上花了100万美元（当然，这会影响你的盈利能力）。","与收入相关的是销售成本（COGS）——即商品销售成本——而对于 AI 来说，COGS 是实实在在的，这与很长一段时间的软件不同。具体来说，在模型上运行推理——无论该模型是 Kimi 还是 Fable——都是有成本的，并且 AI 提供商在推理上花费的金额，至少在大多数商业模式中，通常与收入直接相关。重用上述例子，产生1亿美元与10万美元的收入可能会需要1000倍的 COGS。具体来说，如果生成驱动1美元收入的 token 需要50美分，那么1亿美元的收入将有5000万美元的 COGS；10万美元收入的 COGS 仅为5万美元。","关于开放权重模型的关键点是，它们的服务并非免费。Kimi K3 的成本：https://platform.kimi.ai/docs/pricing/chat-k3 每百万输入 token 3 美元，每百万输出 token 15 美元；这比 Sol 的每百万输入 token 5 美元、每百万输出 token 30 美元要便宜，但这可能甚至不是正确的衡量标准。","英伟达 CEO 黄仁勋将英伟达正在构建的系统描述为“token 工厂”，从英伟达的角度来看，这种表述是合理的。英伟达 GPU 是模型无关的：它们生成 token，并以最快、最有效的方式生成。这就导致了诸如每秒 token 数、生成首个 token 所需时间、每瓦特 token 数、token 成本等指标，黄仁勋认为，这些指标将成为决策的基础。","这种框架在人工智能的第一个范式——ChatGPT时代——中确实是有意义的，那时代币是直接传递给最终用户的。然而，人工智能的第二个范式——推理时代——使这种衡量方式变得复杂。推理意味着链式思维代币数量的爆炸，不同模型需要不同数量的推理代币才能得出正确答案。例如，据报道，Kimi 使用的代币远多于 Sol，从而使其价格优势无效。智能体也引入了类似的动态：在执行智能工作流时，有些模型比其他模型需求的代币更高效。","这意味着代币不是一种商品。商品的定义特征是可互换性：一加仑石油就是一加仑石油；一吨铜就是一吨铜；一蒲式耳小麦就是一蒲式耳小麦。然而，不同模型的代币并不相同。可互换的是由代币构建的东西，也就是智能。换句话说，如果 Kimi 和 Sol 都生成了正确答案，那么该答案是可互换的；为了得到正确答案而生成的代币差异，则会导致生产成本（COGS）的差异。","智能的生产成本（COGS）是由几个不同因素决定的：","这之所以重要，是因为我们正在快速接近这样一种状态：许多经济上有益任务的智能实际上是一种商品。例如，任何构建基本 CRUD 应用的人：https://en.wikipedia.org/wiki/Create,_read,_update_and_delete，很可能可以使用来自多个供应商的模型来完成。而在商品市场中，实现盈利的路径不是通过收取更高的价格——你可以（或者很快就能）使用多个模型制作完全相同的应用——而是通过拥有更优的成本结构。","值得逐步解析这些机制，因为正如我几个月前在《亚马逊的耐久性》中提到的：https://stratechery.com/2026/amazons-durability/，商品市场的动态并不是科技界的人通常所熟悉的。","关键需要理解的是，生产该商品的边际成本因供应商而异。这在实际操作中意味着，拥有最差成本结构的供应商最终会以其边际成本出售商品（如果他们能够生产的话）；其他所有人的利润取决于他们的成本结构相对于边际供应商有多大优势。","假设价格弹性使得在20美元时，该商品的需求量为25单位。这意味着：","这并不是完全正确：供应商C承担短缺的原因是因为供应商A和B能够略低于他们的价格，这当然会影响需求（需求是弹性的），但说明了问题。供应商A有一家很棒的企业，供应商B有一家不错的企业，而供应商C将要破产。","破产风险是固定成本重新成为焦点的地方：供应商C不仅有固定成本（比如可能的研发支出），而且可能承担了债务以资助生产该商品所需的设备。它无法以这些成本来定价商品——记住，市场清算价格大致相当于满足需求所需的最高成本单位的边际成本——但是这些成本绝对可能把供应商逼出市场。如果该供应商倒闭，价格就会上涨，直到另一个供应商决定进入市场（或其他供应商扩张）。","让我们回到模型上。现在，上述分析都不适用，因为前沿模型（frontier models）的需求超过供应，供应受计算能力不足的限制。这种计算能力短缺不仅意味着像Nvidia这样的计算供应商获得非常高的利润率，也意味着Nvidia的客户，如SpaceXAI，可以将计算能力以高利润转售给像Anthropic这样的公司。与此同时，Anthropic可以支付加价，因为他们可以以更高的加价出售代币。","然而，产生Anthropic高利润的不仅仅是过剩的需求：由于模型能力、服务规模和token效率，Anthropic和OpenAI在前沿质量智能的单位成本上可能是最低的。他们在竞争对手之前几个月就开始以特定能力水平服务模型，并同时利用最优模型来优化这些成本。","还值得注意的是，市场尚未将智能视为商品：需求集中在Anthropic和OpenAI，而对不如它们的模型需求要少得多（因此SpaceXAI和Meta将容量出售给Anthropic）；优化成本的动机可以理解为，根据智能水平定义待完成的任务，使智能购买者能够创造一个智能商品化的市场。然而，从长远来看，处于前沿的人也最有可能主导非前沿市场，后者只是前沿市场滞后 n 个月，即前沿模型制造商在这些月份中优化了服务成本。","综上所述，我认为人们对Kimi和中国模型的反应总体上被夸大了，至少从经济角度来看。目前存在一个价格保护伞，这是由于计算资源不足造成的；我非常怀疑中国模型在边际成本基础上更便宜，它们之所以显得便宜，仅仅是因为Anthropic和OpenAI的供应非常紧张，以至于它们收取的价格远高于如果供应充足以满足智能需求时的价格。","那么，为什么模型制造商特别对中国模型如此惊慌？","首先，我认为前沿实验室依然基于训练成本主导其财务模型的世界。当训练消耗的GPU比推理多时，最大化推理收入以资助下一次训练运行至关重要，这意味着要对推理收费非常高。","然而，从现在起，我预计推理市场将增长得比训练成本快得多（这包括假设训练成本将继续飙升），这意味着它们确实可以通过增加量来弥补成本。八个月前，这一点还不很明确，但代理范式的解锁如此庞大，前沿实验室应该更加自信，他们不仅可以生存，还可以在价格降低的情况下繁荣（只要他们拥有足够的计算能力）。","其次，智能事实上并不是一种完美的商品，部分原因是应用智能会使自身变得更智能。具体来说，谁在进行推理，谁也在收集数据，而这些数据会用于下一代模型的改进。一方面，这更是前沿实验室在更多计算能力上线后降低价格、增加使用量的理由；另一方面，这也是为什么像微软这样的公司：https://x.com/satyanadella/status/2076323181154230284 越来越热衷于帮助企业运行自己的模型。如果中国模型是一种可行的替代方案，这将更加可行。","第三，前沿实验室不仅可以通过与中国模型区分开来，还可以通过彼此区分的另一种方式，是继续向上整合到客户体验中。令人惊讶的是，Claude Code 和 Codex 的粘性非常高；你开始使用哪个工具，很可能最终会一直使用它，对于非技术用户来说，这种情况可能更加明显。从长远来看，这种向上整合的迫切需求确实意味着前沿模型对软件提供商，包括微软，是一种威胁。另一方面，目前掌握客户体验的软件公司能获得竞争模型的程度，也决定了它们在多大程度上能够抵御前沿实验室的侵蚀。","最后，尤其是 Anthropic 的意识形态角度：https://stratechery.com/2026/anthropics-safety-superpower/ 不容忽视。这是一家公司，认为只有它可以被信任使用 AI，而开放权重替代品的存在，对这种假设造成了致命打击。","Kimi 并不是唯一的中国新模型；来源：彭博：https://www.bloomberg.com/news/articles/2026-07-19/alibaba-s-qwen-unveils-preview-of-flagship-ai-model","阿里巴巴集团控股有限公司股价周一最高上涨 5.4%，此前公司推出了旗舰模型 Qwen3.8 Max 的预览版本，并称其仅次于 Anthropic PBC 的 Fable 5。周日发布仅几天后，初创公司 Moonshot AI 推出了一个强大的新产品，这动摇了市场，并在美国引发了对中国缩小与 Anthropic 和 OpenAI 等全球领导者差距的担忧。Qwen3.8 Max 拥有 2.4 万亿参数，与 Moonshot 的 Kimi K3 同属重量级类别。K3 拥有 2.8 万亿参数，可与顶级产品竞争，而阿里巴巴也设定了类似的高期望值。","开发者现在可以通过阿里巴巴的编程平台（包括 Qoder）访问 Qwen3.8 Max。阿里巴巴计划很快将该模型的权重开放，使访问不再局限于预览版本。这些国产人工智能系统和模型的兴趣非常高，以至于 Moonshot 被迫在周日晚暂停接受新订阅，以应对压倒性的需求。","Qwen3.8 Max 将开放权重这一事实值得注意。阿里巴巴今年早些时候停止发布其前沿模型的权重，但似乎已恢复这一做法；我怀疑这一变化与上周习近平关于人工智能的讲话有关：http://english.scio.gov.cn/topnews/2026-07/18/content_118605932.html，该讲话强化了开放权重的做法。","我们应坚持开放共赢的原则，推动创新驱动发展。作为世界经济增长的新引擎和增长动力转换的加速器，人工智能正在从数字世界进入物理世界。我们应抓住这一难得的历史机遇，鼓励开源、开放、合作与共享。我们应促进技术创新、产业发展和人工智能的场景化应用。我们应在传统行业转型升级、新兴产业培育成长以及面向未来产业的前瞻布局方面协调推进，使各行各业都能从人工智能中受益。","对中国的策略显而易见：将你的互补品商品化。请注意，习近平明确将开放性与人工智能“从数字世界进入物理世界”联系起来；物理世界是中国主导的世界，而该国在机器人等领域的领先地位将因广泛可用的人工智能模型而大大受益。","沿着同样的思路，中国不希望美国在人工智能方面获得不对称的优势；在中国能够削弱美国前沿实验室的同时加强任何潜在的美国对手时，这对中国更有利，而且它可以从附着在开放生态系统上的创新中受益。","同样，不要指望中国会对前沿实验室的蒸馏攻击采取任何措施。我认为将中国实验室的所有成功归因于蒸馏是错误的，但假装蒸馏没有给中国实验室带来巨大优势也是同样的错误。在过去一年里，随着训练后强化学习对模型性能变得越来越关键，这一优势真正体现出来了。中国实验室无需从零开始设计强化学习环境，只需利用前沿实验室的模型作为教师，就可以以更低的成本实现快速改进（这并不是中国模型开发成本更低的唯一原因，但确实是一个重要原因）。","有趣的是，中国模型在西方的一个最重要的用例本身就是蒸馏。例如，Thinking Machines 刚刚发布了一个开放权重模型，就依赖中国模型：https://www.ft.com/content/ef486929-d2c2-480b-8b00-9cb98bda6acf 来解决强化学习的冷启动问题。Dean Meyer 和 Konstantine Buhler 在 X 上发表了一篇优秀文章：https://x.com/deanmeyerrr/status/2077834267086729674?s=46，解释了蒸馏意味着西方的开放权重模型相对于中国处于根本性劣势：","蒸馏不能解释中国在开放模型方面的全部领先优势。中国实验室拥有世界级的研究人员、充足的算力、强大的预训练模型、软硬件协同设计以及快速提升的后训练能力。但蒸馏压缩了强大基础模型与接近前沿系统之间成本高昂的最终差距。即使蒸馏在中国模型的整体能力中所占比重较小，它在相对于美国开放模型的优势中仍占有重要份额。","新的执法机制将使中国公司进行大规模蒸馏变得更困难、更慢且成本更高。然而，执法不会消除由国家行为者支持的蒸馏。因此，每一次西方前沿进展都为中国实验室创造了另一个教师。西方开发者要么必须独立重现这些能力，要么只能等待从中国模型中学习。这种差距赋予中国实验室相对于西方公司的持续结构性优势。","这一点值得反复强调：由于美国开放权重模型的制造商必须遵守前沿实验室的服务条款，他们（1）比中国的替代品表现更差，且（2）最终只是蒸馏了蒸馏，只是绕道通过中国实验室完成。如果西方开放权重模型制造商能直接获取源头，不是更好吗？","为此，这里有一个关于蒸馏的更有趣的问题：蒸馏到底为什么被认为是坏的？毕竟，大型语言模型不正是开放互联网中所有知识的蒸馏，由前沿实验室抓取后提炼进模型，而这些模型本身又正在被继续蒸馏？到底是谁受到了损害？","整篇文章的目的都是为了缓解人们对 Kimi K3 特别是对中国开放权重模型的过度反应；然而，仍有一个值得关注的原因，那就是网络安全。请参考 The Stack 的报道：https://www.thestack.technology/hugging-face-hacked-turned-to-chinese-llm-for-help-after-us-models-blocked-blue-team/:","Hugging Face表示，其生产基础设施在上周初被一个“自主”AI代理系统入侵。该平台的安全团队表示，他们在事件响应（IR）方面最初遇到阻碍，因为未具名的美国大型语言模型前沿模型护栏“无法区分事件响应者和攻击者”。因此，Hugging Face的防守者转而使用来自中国Z.ai实验室的开源GLM 5.2模型——在他们自己的基础设施上运行它，以分析攻击者留下的17,000多个日志或痕迹。","对于总部位于纽约的Hugging Face来说，这是一个引人注目的公开承认。该公司允许用户在模型、数据集和应用程序上进行协作，并且今年夏天其年经常性收入（ARR）达到了1亿美元。在一份事件报告中，公司建议防御者“在事件发生之前，拥有一个可以在自己基础设施上运行的[斜体为原文强调]可靠模型，并经过审查和准备，以避免护栏锁定，同时防止攻击者的数据和凭证泄露到你的环境之外。”","特朗普政府对Anthropic发布Fable的惊慌反应有多错误，难以言表，尤其是因为这加剧了Anthropic最糟糕的倾向，即认为只有他们值得信任来掌控强大的AI。在一个只有一台AI的世界里，为美国政府及可信盟友保留最强大的网络安全能力可能是有意义的；然而，这并不是我们所生活的世界。","现有的，也将会出现的模型，非常有能力对现有基础设施发起网络安全攻击，而这些模型将会——实际上已经——广泛可用。最好的防御——实际上唯一可行的防御——是确保防御者也能使用最好的模型。目前，由于特朗普政府的指令，防御者实际上被禁止在网络安全中使用Fable或Sol；这意味着最好的替代方案是使用一个多年来一直试图削弱我们网络防御的国家的模型。这简直疯狂！","更好的做法很明确：首先，放宽Fable和Sol在网络安全方面的限制；其次，确保美国的开放重量模型制造商与中国在同一竞争环境中。是的，前沿实验室会对此大为反对，但政府应该意识到，听信他们的过度表演已经导致美国公司在防御上依赖中国。让前沿实验室通过提升自身水平来获胜；不要让他们定义安全或保障，也不要阻碍人类集体知识的进步。中国本身已经很难竞争，让他们引领开放和创新标准，无异于放弃我们最大的优势。","Stratechery Plus：https://stratechery.com/stratechery-plus/","Stratechery历年最受欢迎和最重要的文章。","浏览Stratechery的所有免费文章。","Stratechery Plus：/stratechery-plus/","使用WordPress设计： https://wordpress.com/website-builder/?partner_domain=stratechery.com&utm_source=Automattic&utm_medium=colophon&utm_campaign=Concierge%20Referral&utm_term=stratechery.com 由Pressable托管： https://pressable.com/?utm_source=Automattic&utm_medium=rpc&utm_campaign=Concierge%20Referral&utm_term=stratechery.com"],"translationStatus":"translated","bodyOrigin":"source-page","editorial":{"summary":"Aioga 编辑摘要：中国开源模型 Kimi K3 在能力上逼近当前最先进水平，引发行业讨论。 Aioga 将其归入「技巧观点」方向，重点关注它对真实使用和行业竞争的影响。","background":"背景分析：模型与研究类动态需要结合能力边界、开放方式、成本、可用性和真实任务表现判断，单项指标领先不等于已经形成稳定采用。","viewpoint":"Aioga 判断：这条动态更适合作为行业观察信号，当前信息足以建立线索，但不足以推导长期结论。","implications":"影响分析：对相关团队而言，短期应先核对来源、可用范围和实际成本，再判断是否值得接入或跟进。","nextStep":"后续观察：继续观察官方文档、实际可用性、价格变化、开发者反馈和竞品回应。","evidenceRefs":["title","summary","articleBody"],"confidence":"medium","status":"published","aiGenerated":false,"autoApproved":true,"generatedBy":"rule-safe-fallback","generatedAt":"2026-07-23T06:49:19.094Z","sourceHash":"429aba35215eed83","validation":{"passed":true,"mode":"rule-safe-fallback","checks":["schema","length","source-attribution","no-html"]}},"tags":["技巧观点","Hacker News 热门（buzzing.cc 中文翻译）"],"translations":{"zh-CN":{"title":"谁怕中国模型？--Kimi K3 逼近 SOTA，开源模型成本优势引热议","summary":"中国开源模型 Kimi K3 在能力上逼近当前最先进水平，引发行业讨论。其 API 价格为每百万输入 token 3 美元、每百万输出 token 15 美元，低于 Sol 的 5 美元和 30 美元。但推理时代 token 并非同质化商品，Kimi 需更多推理 token 才能达到正确答案，实际智能成本取决于模型体积、推理效率、内存效率、服务效率和 token 效率等多重因素。","category":"技巧观点","source":"stratechery.com","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"谁怕中国模型？--Kimi K3 逼近 SOTA，开源模型成本优势引热议 - Aioga AI资讯","description":"中国开源模型 Kimi K3 在能力上逼近当前最先进水平，引发行业讨论。其 API 价格为每百万输入 token 3 美元、每百万输出 token 15 美元，低于 Sol 的 5 美元和 30 美元。但推理时代 token 并非同质化商品，Kimi 需更多推理 token 才能达到正确答案，实际智能成本取决于模型体积、推理效率、内存效率、服务效率和 tok...","url":"https://www.aioga.com/news/cmrtwpzse3c55bihzmczi3sbf/"},"en":{"title":"Who’s Afraid of Chinese Models? -- Kimi K3 Approaches SOTA, Open-Source Model Cost Advantages Spark Heated Discussion","summary":"China's open-source model Kimi K3 is approaching the current state-of-the-art level in capability, sparking industry discussion. Its API price is $3 per million input tokens and $15 per million output tokens, lower than Sol's $5 and $30. However, inference-era tokens are not a homogeneous commodity; Kimi requires more inference tokens to reach the correct answer. The actual intelligence cost depends on multiple factors such as model size, inference efficiency, memory efficiency, service efficiency, and token efficiency.","category":"Insights","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Who’s Afraid of Chinese Models? -- Kimi K3 Approaches SOTA, Open-Source Model Cost Advantages Spark Heated Discussion - Aioga AI News","description":"China's open-source model Kimi K3 is approaching the current state-of-the-art level in capability, sparking industry discussion. Its API price is $3 per million input tokens and $1...","url":"https://www.aioga.com/en/news/cmrtwpzse3c55bihzmczi3sbf/","contentTranslated":true,"sourceHash":"3c8da1f3904d386b","translatedAt":"2026-07-22T18:27:10.423Z"},"ja":{"title":"誰が中国のモデルを恐れるのか？--Kimi K3 が SOTA に迫る、オープンソースモデルのコスト優位性が話題に","summary":"中国のオープンソースモデル Kimi K3 は能力において現在最先端レベルに迫り、業界で議論を引き起こしている。その API 価格は、入力トークン100万あたり3ドル、出力トークン100万あたり15ドルで、Sol の 5ドルと30ドルより低い。しかし推論時のトークンは同質のものではなく、Kimi は正しい答えに到達するためにより多くの推論トークンを必要とする。実際の知的コストは、モデルの規模、推論効率、メモリ効率、サービス効率、トークン効率など複数の要因に依存する。","category":"ヒントと視点","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"誰が中国のモデルを恐れるのか？--Kimi K3 が SOTA に迫る、オープンソースモデルのコスト優位性が話題に - Aioga AIニュース","description":"中国のオープンソースモデル Kimi K3 は能力において現在最先端レベルに迫り、業界で議論を引き起こしている。その API 価格は、入力トークン100万あたり3ドル、出力トークン100万あたり15ドルで、Sol の 5ドルと30ドルより低い。しかし推論時のトークンは同質のものではなく、Kimi は正しい答えに到達するためにより多くの推論トークンを必要とする...","url":"https://www.aioga.com/ja/news/cmrtwpzse3c55bihzmczi3sbf/","contentTranslated":true,"sourceHash":"3c8da1f3904d386b","translatedAt":"2026-07-22T18:27:17.190Z"},"ko":{"title":"누가 중국 모델을 두려워하나? -- Kimi K3 SOTA에 근접, 오픈소스 모델 비용 우위로 화제","summary":"중국 오픈소스 모델 Kimi K3는 성능 면에서 현재 최첨단 수준에 근접하여 업계의 논의를 불러일으켰습니다. 그 API 가격은 입력 토큰 100만 개당 3달러, 출력 토큰 100만 개당 15달러로, Sol의 5달러와 30달러보다 저렴합니다. 그러나 추론 시대의 토큰은 동질화된 상품이 아니며, Kimi는 올바른 답을 얻기 위해 더 많은 추론 토큰이 필요합니다. 실제 지능 비용은 모델 크기, 추론 효율, 메모리 효율, 서비스 효율, 토큰 효율 등 여러 요인에 따라 달라집니다.","category":"인사이트","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"누가 중국 모델을 두려워하나? -- Kimi K3 SOTA에 근접, 오픈소스 모델 비용 우위로 화제 - Aioga AI 뉴스","description":"중국 오픈소스 모델 Kimi K3는 성능 면에서 현재 최첨단 수준에 근접하여 업계의 논의를 불러일으켰습니다. 그 API 가격은 입력 토큰 100만 개당 3달러, 출력 토큰 100만 개당 15달러로, Sol의 5달러와 30달러보다 저렴합니다. 그러나 추론 시대의 토큰은 동질화된 상품이 아니며, Kimi는 올바른 답을 얻기...","url":"https://www.aioga.com/ko/news/cmrtwpzse3c55bihzmczi3sbf/","contentTranslated":true,"sourceHash":"3c8da1f3904d386b","translatedAt":"2026-07-22T18:28:03.889Z"},"es":{"title":"¿Quién le teme al modelo chino? -- Kimi K3 se acerca al SOTA, la ventaja de costo de los modelos de código abierto genera debate","summary":"El modelo de código abierto chino Kimi K3 se acerca al nivel más avanzado actual en términos de capacidad, lo que ha generado debate en la industria. Su precio de API es de 3 dólares por cada millón de tokens de entrada y 15 dólares por cada millón de tokens de salida, inferior a los 5 dólares y 30 dólares de Sol. Pero los tokens en la fase de inferencia no son un producto homogéneo; Kimi necesita más tokens de inferencia para alcanzar la respuesta correcta, y el costo real de inteligencia depende de múltiples factores como el tamaño del modelo, la eficiencia de la inferencia, la eficiencia de la memoria, la eficiencia del servicio y la eficiencia del token.","category":"Ideas","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"¿Quién le teme al modelo chino? -- Kimi K3 se acerca al SOTA, la ventaja de costo de los modelos de código abierto genera debate - Aioga Noticias de IA","description":"El modelo de código abierto chino Kimi K3 se acerca al nivel más avanzado actual en términos de capacidad, lo que ha generado debate en la industria. Su precio de API es de 3 dólar...","url":"https://www.aioga.com/es/news/cmrtwpzse3c55bihzmczi3sbf/","contentTranslated":true,"sourceHash":"3c8da1f3904d386b","translatedAt":"2026-07-22T18:28:05.398Z"},"fr":{"title":"Qui a peur du modèle chinois ? -- Kimi K3 approche du SOTA, l'avantage de coût des modèles open source suscite beaucoup de discussions","summary":"Le modèle open source chinois Kimi K3 approche le niveau le plus avancé actuel en termes de capacités, suscitant des discussions dans le secteur. Son tarif API est de 3 dollars pour un million de tokens d'entrée et de 15 dollars pour un million de tokens de sortie, inférieur à celui de Sol, qui est de 5 dollars et 30 dollars. Cependant, les tokens lors du raisonnement ne sont pas des produits homogènes, Kimi a besoin de plus de tokens de raisonnement pour atteindre la bonne réponse, et le coût réel de l'intelligence dépend de plusieurs facteurs tels que la taille du modèle, l'efficacité du raisonnement, l'efficacité de la mémoire, l'efficacité du service et l'efficacité des tokens.","category":"Analyses","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Qui a peur du modèle chinois ? -- Kimi K3 approche du SOTA, l'avantage de coût des modèles open source suscite beaucoup de discussions - Aioga Actualités IA","description":"Le modèle open source chinois Kimi K3 approche le niveau le plus avancé actuel en termes de capacités, suscitant des discussions dans le secteur. Son tarif API est de 3 dollars pou...","url":"https://www.aioga.com/fr/news/cmrtwpzse3c55bihzmczi3sbf/","contentTranslated":true,"sourceHash":"3c8da1f3904d386b","translatedAt":"2026-07-22T18:29:05.372Z"},"de":{"title":"Wer hat Angst vor dem chinesischen Modell? -- Kimi K3 nähert sich SOTA, die Kostenvorteile des Open-Source-Modells sorgen für lebhafte Diskussionen","summary":"Das chinesische Open-Source-Modell Kimi K3 nähert sich in seiner Leistungsfähigkeit dem aktuellen Stand der Technik und löst Diskussionen in der Branche aus. Sein API-Preis beträgt 3 US-Dollar pro Million Eingabe-Token und 15 US-Dollar pro Million Ausgabe-Token, weniger als die 5 US-Dollar und 30 US-Dollar von Sol. Doch Token in der Inferenzphase sind keine standardisierten Güter; Kimi benötigt mehr Inferenz-Token, um die korrekte Antwort zu erreichen. Die tatsächlichen Intelligenzkosten hängen von mehreren Faktoren ab, darunter Modellgröße, Inferenz-Effizienz, Speicher-Effizienz, Service-Effizienz und Token-Effizienz.","category":"技巧观点","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Wer hat Angst vor dem chinesischen Modell? -- Kimi K3 nähert sich SOTA, die Kostenvorteile des Open-Source-Modells sorgen für lebhafte Diskussionen - Aioga KI-News","description":"Das chinesische Open-Source-Modell Kimi K3 nähert sich in seiner Leistungsfähigkeit dem aktuellen Stand der Technik und löst Diskussionen in der Branche aus. Sein API-Preis beträgt...","url":"https://www.aioga.com/de/news/cmrtwpzse3c55bihzmczi3sbf/","contentTranslated":true,"sourceHash":"3c8da1f3904d386b","translatedAt":"2026-07-22T18:28:54.186Z"},"pt-BR":{"title":"Quem tem medo do modelo chinês? -- Kimi K3 se aproxima do SOTA, o modelo de código aberto gera debate sobre vantagem de custo","summary":"O modelo de código aberto chinês Kimi K3 está se aproximando dos níveis mais avançados atuais em termos de capacidade, gerando discussões na indústria. Seu preço de API é de 3 dólares por milhão de tokens de entrada e 15 dólares por milhão de tokens de saída, inferior aos 5 dólares e 30 dólares do Sol. No entanto, tokens na era de inferência não são produtos homogêneos; o Kimi precisa de mais tokens de inferência para alcançar a resposta correta, e o custo real de inteligência depende de vários fatores, como tamanho do modelo, eficiência de inferência, eficiência de memória, eficiência de serviço e eficiência de tokens.","category":"技巧观点","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Quem tem medo do modelo chinês? -- Kimi K3 se aproxima do SOTA, o modelo de código aberto gera debate sobre vantagem de custo - Aioga Notícias de IA","description":"O modelo de código aberto chinês Kimi K3 está se aproximando dos níveis mais avançados atuais em termos de capacidade, gerando discussões na indústria. Seu preço de API é de 3 dóla...","url":"https://www.aioga.com/pt-BR/news/cmrtwpzse3c55bihzmczi3sbf/","contentTranslated":true,"sourceHash":"3c8da1f3904d386b","translatedAt":"2026-07-22T18:29:56.572Z"},"ru":{"title":"Кто боится китайскую модель? -- Kimi K3 приближается к SOTA, обсуждаются преимущества открытых моделей с точки зрения стоимости","summary":"Китайская открытая модель Kimi K3 по своим возможностям приближается к современному передовому уровню и вызвала обсуждения в отрасли. Ее цена API составляет 3 доллара за миллион входных токенов и 15 долларов за миллион выходных токенов, что ниже, чем у Sol — 5 и 30 долларов. Но токены в эпоху рассуждений не являются однородным товаром, Kimi требуется больше токенов для рассуждений, чтобы достичь правильного ответа, фактическая стоимость интеллекта зависит от множества факторов, таких как объем модели, эффективность рассуждений, эффективность памяти, эффективность сервиса и эффективность токенов.","category":"技巧观点","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Кто боится китайскую модель? -- Kimi K3 приближается к SOTA, обсуждаются преимущества открытых моделей с точки зрения стоимости - Aioga Новости ИИ","description":"Китайская открытая модель Kimi K3 по своим возможностям приближается к современному передовому уровню и вызвала обсуждения в отрасли. Ее цена API составляет 3 доллара за миллион вх...","url":"https://www.aioga.com/ru/news/cmrtwpzse3c55bihzmczi3sbf/","contentTranslated":true,"sourceHash":"3c8da1f3904d386b","translatedAt":"2026-07-22T18:29:57.057Z"},"ar":{"title":"من يخاف من نموذج الصين؟ -- Kimi K3 يقترب من SOTA، مزايا التكلفة للنموذج المفتوح المصدر تثير النقاش","summary":"نموذج المصدر المفتوح الصيني Kimi K3 يقترب في قدراته من أحدث المستويات المتقدمة الحالية، مما أثار نقاشات في الصناعة. سعر واجهة برمجة التطبيقات الخاصة به هو 3 دولارات لكل مليون رمز مدخل، و15 دولارًا لكل مليون رمز مخرج، وهو أقل من Sol التي تبلغ 5 دولارات و30 دولارًا. ولكن رموز الحقبة الاستدلالية ليست سلعة متجانسة، يحتاج Kimi إلى المزيد من رموز الاستدلال ليصل إلى الإجابة الصحيحة، وتكلفة الذكاء الفعلية تعتمد على حجم النموذج وكفاءة الاستدلال وكفاءة الذاكرة وكفاءة الخدمة وكفاءة الرموز وعدة عوامل أخرى.","category":"技巧观点","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"من يخاف من نموذج الصين؟ -- Kimi K3 يقترب من SOTA، مزايا التكلفة للنموذج المفتوح المصدر تثير النقاش - Aioga أخبار الذكاء الاصطناعي","description":"نموذج المصدر المفتوح الصيني Kimi K3 يقترب في قدراته من أحدث المستويات المتقدمة الحالية، مما أثار نقاشات في الصناعة. سعر واجهة برمجة التطبيقات الخاصة به هو 3 دولارات لكل مليون رمز م...","url":"https://www.aioga.com/ar/news/cmrtwpzse3c55bihzmczi3sbf/","contentTranslated":true,"sourceHash":"3c8da1f3904d386b","translatedAt":"2026-07-22T18:30:46.583Z"},"hi":{"title":"कौन डरता है चीन के मॉडल से? -- Kimi K3 SOTA के करीब, ओपन-सोर्स मॉडल की लागत लाभ पर गर्म चर्चा","summary":"चीन का ओपन-सोर्स मॉडल Kimi K3 क्षमता में वर्तमान में सबसे उन्नत स्तर के करीब है, जिसने उद्योग में चर्चा को जन्म दिया। इसका API मूल्य प्रति मिलियन इनपुट टोकन 3 अमेरिकी डॉलर और प्रति मिलियन आउटपुट टोकन 15 अमेरिकी डॉलर है, जो Sol के 5 और 30 डॉलर से कम है। लेकिन तर्क प्रयोजन के समय टोकन एकसमान वस्तु नहीं हैं, Kimi को सही उत्तर तक पहुँचने के लिए अधिक तर्क टोकन की आवश्यकता होती है, वास्तविक बुद्धिमान लागत मॉडल के आकार, तर्क दक्षता, मेमोरी दक्षता, सेवा दक्षता और टोकन दक्षता जैसे कई कारकों पर निर्भर करती है।","category":"技巧观点","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"कौन डरता है चीन के मॉडल से? -- Kimi K3 SOTA के करीब, ओपन-सोर्स मॉडल की लागत लाभ पर गर्म चर्चा - Aioga AI समाचार","description":"चीन का ओपन-सोर्स मॉडल Kimi K3 क्षमता में वर्तमान में सबसे उन्नत स्तर के करीब है, जिसने उद्योग में चर्चा को जन्म दिया। इसका API मूल्य प्रति मिलियन इनपुट टोकन 3 अमेरिकी डॉलर और प्रति...","url":"https://www.aioga.com/hi/news/cmrtwpzse3c55bihzmczi3sbf/","contentTranslated":true,"sourceHash":"3c8da1f3904d386b","translatedAt":"2026-07-22T18:30:49.417Z"},"it":{"title":"Chi ha paura del modello cinese? -- Kimi K3 si avvicina al SOTA, il vantaggio di costo dei modelli open source suscita discussioni","summary":"Il modello open source cinese Kimi K3 si avvicina al livello più avanzato attuale in termini di capacità, suscitando discussioni nel settore. Il suo prezzo API è di 3 dollari per milione di token di input e 15 dollari per milione di token di output, inferiore ai 5 dollari e 30 dollari di Sol. Tuttavia, i token nel periodo di inferenza non sono merci omogenee, Kimi ha bisogno di più token di inferenza per ottenere risposte corrette, e il costo effettivo dell'intelligenza dipende da molteplici fattori come la dimensione del modello, l'efficienza dell'inferenza, l'efficienza della memoria, l'efficienza del servizio e l'efficienza dei token.","category":"技巧观点","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Chi ha paura del modello cinese? -- Kimi K3 si avvicina al SOTA, il vantaggio di costo dei modelli open source suscita discussioni - Aioga Notizie IA","description":"Il modello open source cinese Kimi K3 si avvicina al livello più avanzato attuale in termini di capacità, suscitando discussioni nel settore. Il suo prezzo API è di 3 dollari per m...","url":"https://www.aioga.com/it/news/cmrtwpzse3c55bihzmczi3sbf/","contentTranslated":true,"sourceHash":"3c8da1f3904d386b","translatedAt":"2026-07-22T18:31:39.676Z"},"nl":{"title":"Wie is er bang voor het Chinese model? -- Kimi K3 nadert SOTA, open source model kostvoordeel veroorzaakt verhitte discussie","summary":"Het Chinese open-source model Kimi K3 benadert wat betreft capaciteit het huidige geavanceerde niveau en heeft discussies in de sector uitgelokt. De API-prijs bedraagt 3 dollar per miljoen inputtokens en 15 dollar per miljoen outputtokens, lager dan de 5 dollar en 30 dollar van Sol. Maar tokens tijdens het redeneren zijn geen homogene goederen; Kimi heeft meer redeneringtokens nodig om het juiste antwoord te bereiken, en de werkelijke intelligentiekosten hangen af van meerdere factoren zoals modelgrootte, redeneringsefficiëntie, geheugenefficiëntie, service-efficiëntie en talen efficiëntie.","category":"技巧观点","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Wie is er bang voor het Chinese model? -- Kimi K3 nadert SOTA, open source model kostvoordeel veroorzaakt verhitte discussie - Aioga AI-nieuws","description":"Het Chinese open-source model Kimi K3 benadert wat betreft capaciteit het huidige geavanceerde niveau en heeft discussies in de sector uitgelokt. De API-prijs bedraagt 3 dollar per...","url":"https://www.aioga.com/nl/news/cmrtwpzse3c55bihzmczi3sbf/","contentTranslated":true,"sourceHash":"3c8da1f3904d386b","translatedAt":"2026-07-22T18:31:38.355Z"},"tr":{"title":"Kim Çin modellerinden korkuyor?--Kimi K3 SOTA'ya yaklaşıyor, açık kaynak modeli maliyet avantajı tartışmalara yol açtı","summary":"Çin açık kaynak modeli Kimi K3 yetenek açısından mevcut en ileri düzeye yaklaşarak sektör tartışmalarını başlattı. API fiyatı her milyon giriş tokeni için 3 dolar, her milyon çıkış tokeni için 15 dolar olup, Sol’un 5 dolar ve 30 dolarından daha düşüktür. Ancak çıkarım döneminde tokenler homojen mal değildir; Kimi doğru sonuca ulaşmak için daha fazla çıkarım tokenine ihtiyaç duyar, gerçek zekâ maliyeti model boyutu, çıkarım verimliliği, bellek verimliliği, hizmet verimliliği ve token verimliliği gibi birçok faktöre bağlıdır.","category":"技巧观点","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Kim Çin modellerinden korkuyor?--Kimi K3 SOTA'ya yaklaşıyor, açık kaynak modeli maliyet avantajı tartışmalara yol açtı - Aioga AI Haberleri","description":"Çin açık kaynak modeli Kimi K3 yetenek açısından mevcut en ileri düzeye yaklaşarak sektör tartışmalarını başlattı. API fiyatı her milyon giriş tokeni için 3 dolar, her milyon çıkış...","url":"https://www.aioga.com/tr/news/cmrtwpzse3c55bihzmczi3sbf/","contentTranslated":true,"sourceHash":"3c8da1f3904d386b","translatedAt":"2026-07-22T18:32:34.646Z"},"vi":{"title":"Ai sợ mô hình Trung Quốc? -- Kimi K3 tiến gần SOTA, lợi thế chi phí của mô hình mã nguồn mở gây tranh cãi","summary":"Mô hình nguồn mở Trung Quốc Kimi K3 đang tiến gần tới trình độ tiên tiến nhất hiện nay, gây ra nhiều cuộc thảo luận trong ngành. Giá API của nó là 3 USD cho mỗi triệu token đầu vào và 15 USD cho mỗi triệu token đầu ra, thấp hơn so với Sol là 5 USD và 30 USD. Tuy nhiên, token trong thời đại suy luận không phải là hàng hóa đồng nhất, Kimi cần nhiều token suy luận hơn để đạt được câu trả lời đúng, chi phí trí tuệ thực tế phụ thuộc vào nhiều yếu tố như kích thước mô hình, hiệu quả suy luận, hiệu quả bộ nhớ, hiệu quả dịch vụ và hiệu quả token.","category":"技巧观点","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Ai sợ mô hình Trung Quốc? -- Kimi K3 tiến gần SOTA, lợi thế chi phí của mô hình mã nguồn mở gây tranh cãi - Tin tức AI Aioga","description":"Mô hình nguồn mở Trung Quốc Kimi K3 đang tiến gần tới trình độ tiên tiến nhất hiện nay, gây ra nhiều cuộc thảo luận trong ngành. Giá API của nó là 3 USD cho mỗi triệu token đầu vào...","url":"https://www.aioga.com/vi/news/cmrtwpzse3c55bihzmczi3sbf/","contentTranslated":true,"sourceHash":"3c8da1f3904d386b","translatedAt":"2026-07-22T18:32:22.809Z"},"id":{"title":"Siapa yang Takut pada Model Tiongkok? -- Kimi K3 Mendekati SOTA, Keunggulan Biaya Model Open Source Menjadi Perbincangan Panas","summary":"Model open source Tiongkok Kimi K3 mendekati tingkat paling canggih saat ini, memicu diskusi di industri. Harga API-nya adalah 3 dolar per juta token input dan 15 dolar per juta token output, lebih rendah dari Sol yaitu 5 dolar dan 30 dolar. Namun, token pada era inferensi bukan barang homogen, Kimi membutuhkan lebih banyak token inferensi untuk mencapai jawaban yang benar, dan biaya kecerdasan aktual tergantung pada banyak faktor seperti ukuran model, efisiensi inferensi, efisiensi memori, efisiensi layanan, dan efisiensi token.","category":"技巧观点","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Siapa yang Takut pada Model Tiongkok? -- Kimi K3 Mendekati SOTA, Keunggulan Biaya Model Open Source Menjadi Perbincangan Panas - Berita AI Aioga","description":"Model open source Tiongkok Kimi K3 mendekati tingkat paling canggih saat ini, memicu diskusi di industri. Harga API-nya adalah 3 dolar per juta token input dan 15 dolar per juta to...","url":"https://www.aioga.com/id/news/cmrtwpzse3c55bihzmczi3sbf/","contentTranslated":true,"sourceHash":"3c8da1f3904d386b","translatedAt":"2026-07-22T18:33:21.718Z"},"th":{"title":"ใครกลัวโมเดลจีน? -- Kimi K3 ใกล้เคียง SOTA, ข้อได้เปรียบด้านต้นทุนของโมเดลโอเพ่นซอร์สกลายเป็นประเด็นร้อน","summary":"โมเดลโอเพ่นซอร์สของจีน Kimi K3 มีความสามารถใกล้เคียงกับระดับขั้นสูงสุดในปัจจุบัน จนทำให้เกิดการถกเถียงในอุตสาหกรรม ราคา API ของมันอยู่ที่ 3 ดอลลาร์ต่อ 1,000,000 token ขาเข้า และ 15 ดอลลาร์ต่อ 1,000,000 token ขาออก ต่ำกว่า Sol ที่ 5 ดอลลาร์และ 30 ดอลลาร์ แต่ token ในช่วงการคำนวณไม่ใช่สินค้าที่เหมือนกันทั้งหมด Kimi ต้องใช้ token มากขึ้นเพื่อให้ได้คำตอบที่ถูกต้อง ค่าใช้จ่ายเชิงสติปัญญาจริงขึ้นอยู่กับหลายปัจจัย เช่น ขนาดของโมเดล ประสิทธิภาพในการคำนวณ ประสิทธิภาพของหน่วยความจำ ประสิทธิภาพการให้บริการ และประสิทธิภาพของ token","category":"技巧观点","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"ใครกลัวโมเดลจีน? -- Kimi K3 ใกล้เคียง SOTA, ข้อได้เปรียบด้านต้นทุนของโมเดลโอเพ่นซอร์สกลายเป็นประเด็นร้อน - ข่าว AI Aioga","description":"โมเดลโอเพ่นซอร์สของจีน Kimi K3 มีความสามารถใกล้เคียงกับระดับขั้นสูงสุดในปัจจุบัน จนทำให้เกิดการถกเถียงในอุตสาหกรรม ราคา API ของมันอยู่ที่ 3 ดอลลาร์ต่อ 1,000,000 token ขาเข้า และ 15...","url":"https://www.aioga.com/th/news/cmrtwpzse3c55bihzmczi3sbf/","contentTranslated":true,"sourceHash":"3c8da1f3904d386b","translatedAt":"2026-07-22T18:33:28.553Z"},"pl":{"title":"Kto boi się chińskiego modelu? -- Kimi K3 zbliża się do SOTA, zalety kosztowe modelu open source wywołują gorące dyskusje","summary":"Chiński model open source Kimi K3 pod względem możliwości zbliża się do obecnego najbardziej zaawansowanego poziomu, wywołując dyskusje w branży. Jego cena API wynosi 3 dolary za milion tokenów wejściowych i 15 dolarów za milion tokenów wyjściowych, co jest niższe od 5 i 30 dolarów w przypadku Sol. Jednak tokeny w czasie inferencji nie są towarem homogenicznym; Kimi potrzebuje więcej tokenów inferencyjnych, aby uzyskać poprawną odpowiedź, a rzeczywisty koszt inteligencji zależy od wielu czynników, takich jak rozmiar modelu, efektywność inferencji, efektywność pamięci, efektywność usług i efektywność tokenów.","category":"技巧观点","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Kto boi się chińskiego modelu? -- Kimi K3 zbliża się do SOTA, zalety kosztowe modelu open source wywołują gorące dyskusje - Aioga Wiadomości AI","description":"Chiński model open source Kimi K3 pod względem możliwości zbliża się do obecnego najbardziej zaawansowanego poziomu, wywołując dyskusje w branży. Jego cena API wynosi 3 dolary za m...","url":"https://www.aioga.com/pl/news/cmrtwpzse3c55bihzmczi3sbf/","contentTranslated":true,"sourceHash":"3c8da1f3904d386b","translatedAt":"2026-07-22T18:34:17.239Z"}}}}