Kimi K3:开放权重的升级

Kimi K3: The open-weights escalation

内森·兰伯特 Nathan Lambert · Allen Institute for AI · 2026-07-20 · Interconnects ↗

打开互动全文版(逐段中英对照 + 针对本篇提问)→

摘要 · Abstract

本文分析了月之暗面公司发布的 Kimi K3——一个拥有 2.8 万亿参数的开源权重 MoE 模型,及其对 AI 生态系统的影响。文章认为,K3 显著缩小了开源与闭源模型之间、以及中美实验室之间的性能差距,将其从 6-9 个月缩短至 3-5 个月。核心论点是,以月之暗面为代表的中国 AI 实验室不仅仅是依赖蒸馏的快速追随者,而是能够进行前沿创新,这得益于强大的文化和资本效率。文章还讨论了中国对开源 AI 的战略承诺,这体现在习近平在 WAIC 上的主旨演讲中,以及开源模型对闭源实验室的经济减速效应。结论是,虽然开源模型减缓了对前沿实验室的投资,但它们加速了 AI 在经济中的扩散,带来了社会效益,并为解决安全挑战赢得了更多时间,尽管美国仍有望在前沿能力上保持领先。

This article analyzes the release of Moonshot AI's Kimi K3, a 2.8T parameter open-weights MoE model, and its implications for the AI ecosystem. It argues that K3 significantly narrows the performance gap between open and closed models, as well as between Chinese and American labs, reducing it from 6-9 months to 3-5 months. The core thesis is that Chinese AI labs, exemplified by Moonshot, are not merely fast followers relying on distillation but are capable of frontier-level innovation, driven by strong culture and capital efficiency. The article also discusses China's strategic commitment to open-source AI, as signaled by Xi Jinping's WAIC keynote, and the economic decelerationist effect of open models on closed labs. It concludes that while open models slow investment in frontier labs, they accelerate AI diffusion across the economy, offering societal benefits and more time to address safety challenges, though the U.S. is still expected to lead in frontier capabilities.

核心贡献 · Key contributions

局限 · Limitations

章节 · Sections(共 7)

全文 · Full text(逐段中英对照)

对 AI 生态系统的全球影响 The global implications on the AI ecosystem.

7 月 16 日周四,Moonshot AI 发布了其最新旗舰模型 Kimi K3。K3 是一个 2.8T 参数的 MoE 模型,其权重将于 7 月 27 日发布。本文的大部分内容是对生态系统状况的反思,前提是 Moonshot 遵守权重发布日期的承诺。这是对均衡状态的一种更极端的看法,如果情况是中国拥有同样强大但封闭的模型(即 K3 从未发布),许多结果最终会处于中间地带。

On Thursday July 16th, Moonshot AI released their latest flagship model Kimi K3. K3 is a 2.8T parameter MoE model which will have its weights released on July 27th. Much of this article follows as a reflection on the state of the ecosystem, under the assumption that Moonshot keeps their promise of the weights release date. This is a more extreme view of the equilibrium, and many of the results end up in a middle ground if the state of affairs is that China has similarly powerful, but closed models (i.e. K3 is never released).

关键事实是,无论是开源与闭源之间,还是美国与中国之间的模型性能差距,都已从有争议的 6-9 个月缩短到更短的时间,比如 3-5 个月。

The key fact is that either the open-to-closed or American-to-Chinese model performance gap has been reduced from the debated 6-9 months to something shorter, say 3-5 months.

从发布材料来看,K3 显然是一个真正的前沿模型。这将是自 DeepSeek R1 以来开源模型最接近前沿的一次。DeepSeek R1 则是另一回事。那是一家中国实验室极快地转向推理模型,并比许多美国公司更快地发布了一款推理模型。Kimi K3 是中国实验室在已知领域(数据、算法、架构、工具、环境等)执行 Scaling(规模扩张)的一个例子。

From the release materials, it is clear that K3 is a true frontier model. It will be the closest open models have been to the frontier since DeepSeek R1. DeepSeek R1 was a different story. This was a Chinese lab being extremely quick to pivot to reasoning models and release one faster than many American companies. Kimi K3 is an example of a Chinese lab executing on scaling the known areas: data, algorithms, architecture, tools, environments, etc.

Kimi K3 在 Vals AI 指数上总体排名第二,在 Artificial Analysis 的智能指数上总体排名第三(仅被 Claude Fable 和 GPT-5.6 Sol Max 击败,但价格更便宜),在前端代码竞技场中总体排名第一,还有更多令人印象深刻的结果。Moonshot AI 正以远少得多的资源与 Anthropic 和 OpenAI 正面交锋。

Kimi K3 comes in at #2 overall on the Vals AI index, #3 overall on Artificial Analysis’s Intelligence Index (only beaten by Claude Fable and GPT-5.6 Sol Max while being cheaper), #1 overall in Frontend Code Arena, and more impressive results. Moonshot AI is going toe to toe with Anthropic and OpenAI with far, far fewer resources.

这显然是有史以来发布的最强开源模型。观察这个模型应该清楚,如果来自美国封闭前沿模型的对抗性蒸馏有所贡献,那至多也是相对较小的程度。那些追随蒸馏恐慌并得出错误结论——认为中国 AI 实验室仅因知识产权盗窃才能生产好模型——的 AI 观察者们将幡然醒悟:中国公司非常擅长构建模型,就像领先的美国公司一样。Moonshot AI 正在解决许多 OpenAI 或 Anthropic 的人们正在解决的相同问题。我相信会有更多关于蒸馏的讨论和压力,但证据已经表明,中国公司能做的不仅仅是快速跟随。

It is clearly the strongest open model ever released. It should be clear looking at this model that if adversarial distillation from the closed frontier models in the U.S. contributed, it is at most to a relatively small degree. AI observers who followed the distillation panic and came away with the wrong conclusion that Chinese AI labs are only producing good models due to IP theft are in for an awakening – that Chinese companies are extremely good at building models in the same way the leading American companies are. Moonshot AI is solving many of the same problems that folks at OpenAI or Anthropic are solving. I’m confident there will be more distillation discussion, and pressure, but the evidence is now out that Chinese companies can do more than just fast following.

在中国之行中,我与 Kimi 核心团队的部分成员会面,我清楚地感受到他们拥有非凡的文化——有人会称之为气场——以及在 GPU 受限环境中自由表达这种文化的能力。在构建模型很大程度上是一场 Scaling(规模扩张)游戏的领域里,构建优秀模型的能力在很大程度上仍取决于个体的执行力、动机与表达。拜访过他们之后,这一结果就不那么令人意外了。我拜访过许多 AI 公司,很少有公司能让你立刻感受到这样的文化。

Meeting some of the core Kimi team on my trip to China, it was clear to me that they had incredible culture, some would say aura, and a freedom to express it – within the constraints of a GPU-limited environment. Where building models is so much of a scaling game, much of the ability to build a good model still comes down to individual execution, motivation, and expression. Having visited them, this result is less surprising. Having visited many AI companies, very few have a culture that you can immediately pick up like this.

与此同时,中国的 AI 采用趋势起步晚于美国。因此,尽管所有中国实验室的算力都远少于美国同行,但其中更多算力无疑可以用于训练。当我开玩笑说 OpenAI 一位普通研究员可能拥有多少算力——比如几千台 H100 等效机器——Kimi 的研究人员都感到震惊。Kimi 模型的组织架构与构建方式无疑反映了这一点,但若没有大量专有信息,很难厘清其具体面貌。

At the same time, China's AI adoption trends started later than those in the U.S. So, while all the Chinese labs have way less compute than their counterparts in the U.S., more of it can certainly go to training. When I joked around about how much compute an average researcher at OpenAI could have – say a few thousand H100 equivalent machines – the researchers at Kimi were shocked. The org chart and approach to building the Kimi models surely reflect this, but it is difficult to tease out what this looks like without substantial proprietary information.

关于峰值模型性能的现状大致如下:

The state of affairs on peak model performance is roughly as follows:

3. Moonshot AI – Kimi K3(开放权重*)

3. Moonshot AI – Kimi K3 (open weights*)

5. 智谱(Z.ai)– GLM 5.2(开放权重)

5. Zhipu (Z.ai) – GLM 5.2 (open weights)

8. 阿里巴巴——Qwen 3.7 Max(已宣布 3.8 版本,在撰写本文时也将采用开放权重)

8. Alibaba – Qwen 3.7 Max (3.8 announced, also to be open-weights, at the time of writing)

看到 DeepMind 以及其他一些美国巨头排名如此之低,令人惊讶。在许多方面,xAI 团队值得更多赞誉。下面是来自 Artificial Analysis 的可视化总结:

It is astonishing to see DeepMind, and some of the other American giants, this low. In many ways, the xAI team deserves more credit. A visual summary from Artificial Analysis is below:

在许多方面,这感觉像是一个新时代的开端。一个竞争激烈得多的时代,但同时也对协调提出了高得多的需求,因为我们正在全球范围内部署极其强大的技术。

In many ways, it feels like the start of a new era. An era with much more competition, but also a much higher need for coordination, as we roll out incredibly powerful technologies around the world.

1. 中国重申对开源 AI 的承诺——对近期风险的不同解读 1. China’s recommits to open-source AI – showing a different read on near-term risks

许多人最近才开始关注中国的人工智能领域,因此他们可能会得出结论,认为公开模型是中国的核心战略。事实上,我认为大多数实验室的核心战略更接近 Anthropic 或 OpenAI——构建尽可能最好的智能。多年来,我一直关注并与中国实验室互动,对于他们最初转向公开模型的最佳解释是实用性。他们需要公开模型来获得采用、关注和反馈(尤其是在高价值的湾区市场)。

Many people started following China's AI scene relatively recently, so they can reach the conclusion that releasing models openly is their core strategy. In fact, I think most labs have a core strategy far closer to Anthropic or OpenAI – build the best intelligence possible. Having followed and engaged with the Chinese labs for years now, the best explanation for their original turn to releasing their models openly is practicality. They needed to release the models openly to get adoption, attention, and feedback (especially in the high-value, Bay Area market).

很长一段时间以来,中国在解释开源人工智能的作用以及什么是“国家层面战略”方面的政策非常有限。据我所知,没有高级领导公开评论过开源人工智能。本周情况发生了变化,习近平在世界人工智能大会(WAIC)上发表了主旨演讲,非常直接地将中国人工智能生态的未来承诺给开源和全球扩散。对现状的这一承诺,与迄今为止最强的开放权重模型的宣布在同一周,是现代人工智能早期历史中的一个清晰标志。

For a long time, there had been very limited policy in China explaining the role of open-source AI, and what could be the "country-level strategy." To my knowledge, no senior leaders had commented on open-source AI publicly. This changed this week too, as Xi Jinping gave a keynote address at the World AI Conference (WAIC), and very directly committed the future of China's AI ecosystem to open-source and global diffusion. This commitment to the status quo, the same week as the announcement of the strongest open-weight model to date, is a clear mark in the early history of modern AI.

这发生在一个为中国人工智能产业讨论了许多潜在前进道路的时期——他们会保持开放吗?他们能在 Scaling(规模扩张)方面跟上美国实验室吗?中国有一个不断增长的营收市场吗?在这些问题中,焦点一直是中国对风险的容忍度、公司的变现能力,以及公司停止公开其_最佳_模型的任何密切相关的理由。

This comes during a time period where many potential paths forward have been discussed for the Chinese AI industry – Will they stay open? Can they keep up with the American labs in scaling? Is there a growing revenue market in China? With these, the focus has been on China's risk tolerance, the companies' ability to monetize, and any closely related reason for a company to stop releasing their _best_ models openly.

将习近平的承诺与一个非常强大的模型在时间上联系起来,中国隐含地评论了其在发布开放权重模型方面的风险容忍度。就目前而言,这是对中国系统内对强大网络安全能力(或生物危险)等主题的感知风险的解读。

In tying Xi's commitment in time to a very strong model, China has implicitly commented on its risk tolerance with respect to releasing open-weight models. For the time being, it is a read into the perceived risks of topics like strong cybersecurity capabilities (or bio-dangers) within the Chinese system.

最简单的解释是,中国政府肯定在密切关注模型的潜在风险——可能比美国政府的“氛围监管”具有更多的技术范围——并且如果测量到风险就会采取行动。简单的解释是,他们认为当前的前沿模型没有有意义的风险。

The simplest explanation is that China's government is definitely following potential risks from the models closely – likely with more technical scope than the US government's vibe regulation – and would take action if it measured risk. The simple explanation is that they do not find current frontier models to have meaningful risk.

与此同时,中国的经济决策者认为推动 AI 应用是好事,这样他们可以稍后从该产业中获利——在扩大分发规模之后(正如中国近年来在汽车、太阳能、先进制造以及许多领域所做的那样)。

At the same time, China's economic decision makers think having AI adoption is good, so they can make profits on the industry later – after growing distribution (as China has done for cars, solar, advanced manufacturing, and many areas in recent history).

在美国 AI 媒体界经历了数月围绕 Claude Mythos 模型的炒作与恐慌渲染之后,这些做法可能显得有些令人震惊。这种意外应当成为极好的立足点——世界对于我们在美国最常听到的那些关于 AI 的叙事,并没有一致的共识。

These can seem somewhat shocking, in an American AI media landscape that has gone through months of hype and fearmongering over the Claude Mythos model. This surprise should be excellent grounding – the world does not have a unanimous agreement with the narratives about AI that we hear most in the U.S.

2. 开放模型作为前沿实验室的经济阿喀琉斯之踵 2. Open models as the economic Achilles heel of frontier labs

许多引导人工智能讨论的叙述者有明确的动机去压低最佳开放人工智能模型被感知的能力。Dean Ball——他个人支持开放模型,但现在在 OpenAI 工作——发表了一篇被广泛评论的文章,其中包含对 Kimi 的一些思考,他在文中对开放模型说了以下的话。理解这一陈述很重要,因为它聚焦于开放模型在人工智能建设的经济层面的作用。Dean 说:1

Many of the narrators guiding the discussion on AI have clear incentives to depress the perceived capabilities of the best, open AI models. Dean Ball – who is personally supportive of open models, but now works at OpenAI – had a widely commented on post with some reflections on Kimi, where he said the following on open models. It is important to understand the statement, as it focuses the role of open models in the economic side of the AI buildout. Dean says:1

1. 开放权重模型本质上是减速主义的,我一直惊讶地看到所谓的“加速主义者”对开放权重模型如此兴奋。

1. Open-weight models are inherently decelerationist, and I’m continually surprised to see the so-called “accelerationists” so excited about open-weight models.

解释为什么开放权重模型是_一种减速主义形式_对于理解即将到来的世界秩序很重要。他是对的。

Explaining why open-weight models are _a form of decelerationism_ is important to understanding the coming world order. He is right.

开放模型在经济上对前沿实验室是减速主义的,这将减缓人工智能的净投资和资本支出推出。这是因为强大的开放权重人工智能模型大幅降低了封闭实验室的利润潜力。这有两个效应。第一,人工智能实验室有更少的利润再投资于未来模型。第二,市场认为这些公司的终值更低,因此它们将削弱未来的融资轮次。这些加在一起将减缓最变革性人工智能模型的时间线,但我并不认为这些效应强大到足以阻止 OpenAI 和 Anthropic 成为世界上少数几家估值最高的公司。

Open models are decelerationist economically for the frontier labs, which will slow the net investment and capex rollout for AI. This is due to the fact that strong open-weight AI models massively reduce the margin potential for the closed labs. This has two effects. First, the AI labs have fewer profits to re-invest into future models. Second, the market sees the terminal value of these companies as being lower, so they will kneecap future fundraising rounds. These together will slow timelines to the most transformative AI models, but I do not see them as strong enough effects to stop OpenAI and Anthropic from being a few of the top valued companies in the world.

对我来说,这些对社会是净利好。因为开放权重模型通过降低特定性能水平智能的入门价格,对_人工智能在经济中的扩散_是加速主义的。开放模型也鼓励定制化。问题是,这种扩散本质上比前沿人工智能实验室的产品慢得多,后者销售直接由开发者使用的工具。开放模型的潜力在于几乎每个企业都能使用它们来打造领域特定的智能体。这种经济扩散需要极长的时间!我将其描述为开放权重模型起步慢得多,但潜在指数更大。问题在于,如果封闭模型在原始能力上领先太多,这种定制化的能力可能变得无足轻重。

These, to me, are a net good for society. As open-weight models are accelerationist for _AI diffusion across the economy_ by having the entry price for intelligence at a certain level of performance be lower. Open models also encourage customization. The thing is that this type of diffusion is by its nature far slower than the frontier AI labs products, who sell tools used directly by developers. The potential for open models is for nearly every business to use them to craft domain-specific agents. This economic diffusion takes an extremely long time! I’ve described this as open-weight models being on a much slower starting, but potentially bigger exponential. The problem is, if closed models get too far ahead in raw capabilities, this ability to customize can be moot.

我认为,AI 实验室中扩散增强与权力集中度下降的结合,对 AI 转型非常有利。它给了我们更多时间去弄清楚新能力带来的难题,并让更多利益相关者影响这一进程——任何一家公司都很可能在安全控制世界上最重要的技术方面遇到问题。

The combination of increased diffusion and decreased concentration of power in the AI labs I see to be very positive for the AI transition. It gives us more time to figure out the hard problems of new capabilities and lets more stakeholders impact the story – any one company is very likely to have issues with controlling the world's most important technology safely.

当然,在这个世界中,最好的模型仍由美国公司制造,对我来说很重要,这将使美国能够控制这项技术的轨迹及其价值观。我也预计情况会如此,因为美国拥有更大的资本市场,愿意投资 AI(以及不断增长的利润份额),但这并非理所当然。

It is, of course, important to me in this world for the best models to still be made by U.S. companies, which will allow the U.S. to control the trajectory of the technology and its values. I also expect this to be the case, as the U.S. has larger capital markets that are willing to invest in AI (and a growing share of profits), but it is not a given.

Interconnects AI 是一份由读者支持的出版物。考虑成为订阅者。

Interconnects AI is a reader-supported publication. Consider becoming a subscriber.

3. 中国的效率优势 3. China’s efficiency advantage

Kimi 的发布博客包含一些技术细节,证实了那些为我们所感受到的持续模型改进提供支持的改进类型。仅举一例:

Kimi's launch blog has some technical details that confirm the sort of improvements that are supplying the consistent model improvements we feel. To select one:

Kimi K3 基于 Kimi Delta Attention(KDA)和 Attention Residuals(AttnRes)构建,这两项架构更新旨在改善信息在序列长度和模型深度上的流动方式。我们还扩大了混合专家(MoE)的稀疏性,在搭配 Stable LatentMoE 框架时,有效激活了 896 个专家中的 16 个。结合精炼的训练和数据配方,这些结构性变化相比 Kimi K2 在整体 Scaling(规模扩张)效率上带来了约 2.5 倍的提升,使模型能够更有效地将算力转化为智能。

Kimi K3 is built on Kimi Delta Attention (KDA) and Attention Residuals (AttnRes), two architectural updates designed to improve how information flows across sequence length and model depth. We have also scaled up Mixture of Experts (MoE) sparsity, effectively activating 16 out of 896 experts when paired with a Stable LatentMoE framework. Together with refined training and data recipes, these structural changes yield an approximate 2.5× improvement in overall scaling efficiency compared to Kimi K2, allowing the model to convert compute into intelligence more effectively.

训练效率确实会累积起来。它们将为模型带来持续且惊人的进步。

Training efficiency really adds up. They will result in continued, incredible steps for the models.

顺便说一句,追踪这一特定创新在生态系统中的路径是一个有趣的例子。Kimi Delta Attention(KDA)在 Kimi Linear 论文中被提出,类似于用于 Olmo Hybrid(我在 Ai2 时最后一个 Olmo 模型)的 Gated DeltaNet。Qwen 的最新模型转向了相关架构,最近的 Nemotron 模型也是混合架构(但仍更接近 Mamba 而非 Gated DeltaNet)。看到这些新的架构思想——它们在学术界得到了大量推进——如此迅速地转化为前沿规模的模型,真是令人惊叹。Gated Delta Networks 于 2024 年底提出,建立在 Mamba 的思想之上。到 2026 年中期,它们已应用于前沿模型。

As an aside, tracing the path of this particular innovation through the ecosystem is an interesting example. Kimi Delta Attention (KDA) was introduced in the Kimi Linear paper, which is similar to the Gated DeltaNet used for Olmo Hybrid (my last Olmo model while at Ai2). Qwen's latest models switched to a related architecture and the recent Nemotron models also are hybrid (but still closer to Mamba than Gated DeltaNet). It's awesome to see new architecture ideas like these, which were heavily progressed by academia, get so quickly translated into frontier-scale models. Gated Delta Networks were introduced in late 2024, building on ideas from Mamba. By mid 2026, they're in frontier models.

我选择聚焦这个例子,部分是因为 Kimi 团队为模型间的创新给出了一个很酷的数字,但主要是为了给更广泛地讨论中国的资源效率留出空间。

I chose to focus on this example, partially because the Kimi team put a cool number to innovations between models, but primarily to give space to a broader discussion of China's resource efficiency.

越来越明显的是,中国实验室的资本效率要高得多。在缩放定律决定智能与有效资本(用于购买算力、数据和人才)成正比的世界里,这可能是贵国人工智能产业所能拥有的最大优势。造成这种情况的可能解释有很多,例如中国研究人员薪酬较低,但在解决大语言模型研究难题上更有效率,但我们可能永远无法获得如此具体的原因。

It is becoming clear that the Chinese labs are far more capital efficient. In a world where scaling laws dictate that intelligence is proportional to effective capital – which buys compute, data, & talent – that may be the greatest strength your AI industry could ever have. There are many possible explanations for why this is the case, such as Chinese researchers being paid less while being more effective at LLM research puzzles, but we will probably never get such specific reasons.

自从我写下关于中国的笔记以来,我听到更多关于中国新兴数据产业的消息(远远落后于 Anthropic 在数据上的十亿美元预算),以及中国实验室能够获得有意义的训练算力(通过规避出口管制)。中国公司_没有_同样的推理需求(直到最近,Moonshot AI 不得不暂停其 K3 模型的新订阅访问——而 API 仍然在线),因此它们更多的算力可以用于训练。这些领域严重影响着实验室能力的真实性,但我们对它们的测量非常有限。

Since writing my notes on China, I’m hearing more about an emerging data industry in China (far behind the billion dollar budgets of Anthropic for data) and that Chinese labs have access to meaningful training compute (by skirting export controls). Chinese companies _do not_ have the same inference demand (until recently, as Moonshot AI had to pause new subscriptions for access to their K3 model - while the API is still live), so much more of their compute could go to training. These areas impinge heavily on the truth of the ability of the labs, but we have very limited measurement into them.

实际情况是,这些中国实验室筹集的资本比美国人工智能生态系统的任何部分都少几个数量级。最直接的比较是 OpenAI 和 Anthropic,它们拥有稍好的公开模型。其他公司,如谷歌和 Meta,拥有商业史上最大的现金流,但在构建模型方面落后。至于美国的新实验室,情况更具竞争力——Thinking Machines 最近发布了他们的第一个模型 Inkling,它很强,但无法与 Kimi K3 相提并论。

The facts on the ground are that these Chinese labs have raised orders of magnitude less capital than any slice of the American AI ecosystem. The most direct comparisons are to OpenAI and Anthropic, who have slightly better public models. Others, such as Google and Meta have the largest cash flows in the history of business, and are behind on building models. As for American neolabs, the picture is even more competitive – Thinking Machines released their first model recently, Inkling, which is strong but not in the same class as Kimi K3.

这些拥有更多资源的美国公司仍可能迎头赶上,但你需要强烈权衡我们拥有的模型质量的公开测量结果,而不是诉诸希望——希望往往反映了一种偏见。如果中国实验室确实拥有潜在优势,他们可以继续利用这一点来构建比_所有_竞争对手更强大的模型!许多结果是可能的,K3 应该会增加大多数人对中国在不久的将来凭借更高效的训练努力在人工智能能力上 outright 领先的概率——即使这不是你最可能预测的结果。

These American companies with more resources could still catch up, but you need to strongly weigh the public measurements we have of model quality and not resort to hope – which often reflects a bias. If the Chinese labs do have a latent advantage, they could continue to utilize that to build even stronger models than _all_ the competitors! Many outcomes are plausible and K3 should increase most people’s probability that China can outright lead in AI capabilities in the near future on the back of more efficient training efforts – even if it’s not your most likely predicted outcome.

资本效率的一个主要贡献者可能在于方法上,美国实验室花费大量精力以戏剧性的大步推动前沿,而中国实验室更专注于追赶——这种追赶更便宜。正如学生模型在蒸馏中通常可以超越教师模型(不限于中国实验室的_对抗性_蒸馏),一种“试图追赶”而非“发明下一个范式”的方法可能会产生更强大的模型。

A big contributor to the capital efficiency is likely in the approach, where American labs are spending meaningful energy in pushing the frontier in dramatic, big steps, and the Chinese labs are more focused on catching up — this catch-up is cheaper. Just as the student model can outperform the teacher in distillation generally (not limited to the _adversarial_ distillation of the Chinese labs), an approach of “trying to catch up” rather than “invent the next paradigm” could lead to stronger models.

4. 前沿开放模型日益壮大的生态系统 4. A growing ecosystem of frontier, open models

在 Kimi K3 发布后的那个周末,笔者在撰写本文并广泛讨论相关事件时,阿里巴巴宣布即将推出一个 2.4 万亿参数的 Qwen 3.8 模型,并开放权重。历史上,阿里巴巴一直将其最大的模型通过云业务以仅提供 API 的形式提供,因此这是又一次重大的氛围转变,为开放模型经济的下一篇章打开了大门。即使该模型在基准测试上落后于 Kimi K3,这也表明中国公司可能不仅是在维持其开放模型战略的现状,而是在进一步加码。

The weekend after the Kimi K3 release, while writing this and discussing the events broadly, Alibaba announced that a 2.4 trillion parameter Qwen 3.8 model is coming soon with open-weights. Historically, Alibaba has kept their largest models as API-only offerings via their cloud business, so this is another big vibe shift opening the doors to the next chapter of the open model economy. Even if the model is behind Kimi K3 on benchmarks, it signifies that Chinese companies may not only be maintaining the status quo for their open model strategy, but leaning further into it.

如果这个 Qwen 3.8 模型很快发布,即在下一个 Gemini 模型之前发布,它可能将 Google 推到最智能模型实验室排行榜的第 8 位——而中国在这份榜单上一直在攀升。还有其他传言称,很快会有更多强大的中国模型出现,其中 DeepSeek V4 预计将脱离其“预览”版本。

If this Qwen 3.8 model releases soon, i.e. before the next Gemini model, it could push Google to the 8th position on the leaderboard of labs with the smartest models – a list that China has been climbing. There are other rumors of more strong Chinese models soon, with DeepSeek V4 expected to graduate out of its "preview" version.

5. 前沿开放权重政策漫长故事的开端 5. The very beginning of a long story of frontier open-weight policy

我认为,如果 Claude Mythos 今天作为开放权重模型发布,负面后果会相对较小。这是一个多少有些难以坚持的观点,因为公开的网络安全评估非常有限,生态系统也很复杂(也因为我相信 Anthropic 的许多人)。我仍然坚持这一观点。风险被过度炒作了。

I think that if Claude Mythos was released as an open-weight model today, the negative outcomes would be relatively minor. This is a somewhat challenging opinion to hold, as we have very limited public cybersecurity evaluations and it is a complicated ecosystem (and because I trust many people at Anthropic). I still stand by it. The risks have been over-hyped.

问题在于,对于最强的 AI 模型而言,情况不会总是如此。强大得多的模型正在到来——随之而来的是更高的风险——因此,让最好的模型在类似开放权重模型出现之前数月内以受控、封闭的方式被访问,是一种极其安全的均衡。

The problem is that this will not always be the case for the strongest AI models. Far stronger models are coming — and with them increased risks — so it is an incredibly safe equilibrium for the best models to be accessed in a controlled, closed manner several months ahead of similar open-weight models.

用户高度可控的开放权重模型总会不断出现——你无法有效禁止数字产品,尤其是来自恶意行为者的产品——因为 AI 训练被证明在全球范围内的可及性比许多分析师预期的持续得更久。

Open-weight models which are very controllable by the user will always be coming — you cannot effectively ban digital products, especially from bad actors — as AI training has proven globally accessible longer than many analysts expected.

尽管如此,在我写这篇文章时,美国政府仍在试探更多旨在限制开放权重模型的措施。最新消息来自 Axios:

Still, as I write this, the government continues to flirt with more measures aimed at restricting open-weight models in the U.S. The latest is from Axios:

幕后消息:一位接近政府的消息人士告诉 Axios,美国商务部去年曾考虑将多家中国 AI 实验室列入其“实体清单”,这将实际上在没有许可证的情况下切断美国的访问。

Behind the scenes: The Commerce Department last year considered adding multiple Chinese AI labs to its “Entity List,” which would effectively cut off U.S. access without a license, a source close to the administration told Axios.

* 消息人士称,美国国家安全局和白宫国家网络主任办公室去年也曾考虑就中国 AI 实验室的威胁发布一份咨询报告,实际上是在劝阻美国公司使用它们的技术。

* The National Security Agency and the White House Office of the National Cyber Director also considered issuing an advisory on threats from Chinese AI labs last year, effectively discouraging U.S. companies from using their technology, the source said.

* 该消息人士补充说,白宫曾考虑实施一项行政命令,规定美国公司只有在能够保证安全并承担违规责任的情况下,才能托管中国模型。

* The White House considered implementing an executive order stating that U.S. companies could only host Chinese models if they could guarantee security and assume liability if a breach occurred, the source added.

* 另一位接近政府的消息人士表示,去年夏天,商务部还在政府内部传阅了规则草案,利用其确保国内供应链安全的权力来针对中国的开源模型。

* Last summer, the Commerce Department also circulated draft rules within the administration leveraging its authorities to secure domestic supply chains to target Chinese open-source models, another source close to the administration said.

这将使美国处于一个非常不对称的状态:美国最好的模型在网络安全任务上设有护栏,但全球行为者却可以获取优秀的中国开放权重模型来探测我们的防御。这只是众多例子之一,说明禁止开放权重模型不仅损害 AI 的自由市场,而且在短期内使生态系统更不安全。还有其他非常糟糕的后果,比如减缓 AI 应用和 AI 研究的扩散,正如我上面所讨论的。

This would leave the U.S. in a very asymmetric state where the best models in the U.S. have guardrails on cybersecurity tasks, but global actors have access to great Chinese open-weight models to probe our defenses. This is one of many examples where banning open-weight models not only harms the free markets of AI but also makes the ecosystem less safe in the short term. There are other very bad outcomes, such as slowing the diffusion of AI applications and AI research, as I discussed above.

这些平衡很难维持,尤其是当 AI 工具加速模型进步时,但保持开放与封闭之间的现状很重要。拥有一个在能力上真正独处前沿的模型——比如 Mythos 刚发布时那样——同时也是开放权重的,这在我们进入未知能力领域时构成严重风险。模型将进步得非常快,衡量它们的总能力越来越困难。

These equilibria are very hard to maintain, especially as AI tools accelerate progress in models, but it is important to maintain this status quo between open and closed. Having a model that is truly alone at the frontier in capabilities — something like Mythos when it was announced — also be open-weight poses serious risks as we venture into the unknown of capabilities. Models are going to progress very fast and it is increasingly hard to measure their total capabilities.

于是我们陷入这样一种处境:我们试图在开放模型问题上做到恰到好处。不希望如此强大的东西在一瞬间扩散到全球,这是合理的;但与此同时,模型的制造者有动机去夸大其能力,而它们的竞争对手则有动机去夸大其风险。归根结底,这一切都取决于审慎的测量,以及主动增强社会对各类风险向量的抵御能力。

We are then stuck in a world where we are trying to thread the needle on open models. It's reasonable to not want something so powerful to be diffused globally in an instant, but meanwhile the makers of the models are incentivized to hype their capabilities, and their competitors are incentivized to hype their risks. It all comes down to careful measurement and proactive hardening of society to risk vectors.

这趟由政策辩论、模型发布和喧嚣反应交织而成的疾驰列车,从今天起只会继续下去。自去年十二月 Claude Opus 4.5 发布、将我们带上智能体式路径以来,我们一直处在快速进展的轨道上,AI 应当如何发展的所有关键理念都在此接受检验。在此做出良好决策的关键,是独立于那些财务利益最大的公司的评估能力。因此,随着我们进入 AI 治理的 AGI 时代,所需的众多行动之一,是采取类似“曲速行动”(Operation Warp Speed)的方式,扶持国家能力(以及其他独立行为体),使其能够准确评估模型并研究新兴风险。

This careening train of policy debates, model releases, and raucous reactions is only going to continue from today. We've been on a train of rapid progress, where all the key ideas of how AI should play out are tested, since the release of Claude Opus 4.5 last December, which sent us down the agentic pathway. The key to making good decisions here is evaluation capabilities, independent of the companies with the largest financial stakes. One of many actions needed then, as we enter the AGI era of AI governance, is an Operation Warp Speed style approach of bootstrapping state capacity (and other independent actors) that can evaluate models accurately, and study emerging risks.

结论:警钟 Conclusion: The wake-up call

开放权重模型通过加速能力的扩散,在 AI 的益处和潜在危害方面都带来了巨大的升级。目前,前沿语言模型的负面影响在很大程度上仍是假设性的,但这种情况不会永远持续。

Open-weight models, by accelerating the diffusion of capabilities, are a massive escalation in the good and the potential bad of AI. For now, the bad side of frontier language models has been largely hypothetical, but that will not always remain the case.

让开放权重模型略微落后于封闭前沿,是我们缓解风险的自然缓冲。关键在于,我们必须共同行动,在潜在危害出现时加以缓解,无论开放权重模型落后 3 个月、6 个月还是 9 个月,那仍然是非常短的时间线。如果我们对开放权重模型进行严厉监管,我怀疑世界大部分地区会被麻痹,认为我们不再需要行动。我们所做的只是稍微推迟了不可避免之事——开放模型最终将继续跨越所有关键能力阈值,无论合法与否。

Having open-weight models be slightly behind the closed frontier is our natural buffer to mitigate the risks. The key point is that we must collectively act to mitigate potential harms as they appear, and whether open-weight models are 3 or 6 or 9 months behind, that is still a very short timeline. If we regulate open-weight models heavy-handedly, I suspect much of the world will be lulled into thinking we no longer need to act. All we would've done is slightly delayed the inevitable — open models will continue to cross all the key capability thresholds eventually and regardless of legality.

理解并从这种开放-封闭的舞蹈中获益,必须是 AI 社区在未来几年内跨越所有权力和影响力部门的集体行动。

Understanding and benefiting from this open-closed dance must be a collective action from the AI community across all sectors of power and influence over the coming years.

Kimi K3 是一个分水岭时刻,因为前沿开放权重模型现在已成为现实。许多关于开放权重模型风险真正落地的假设将得到检验——我怀疑它会比许多人预期的更窄,而 AI 的许多风险仍将通过封闭且“更安全”的 API 扩散。对这些风险的评估将随着时间的推移而演变,并随着 AI 在我们经济中整合的加速而加速。我们无法只得到其中一个而没有另一个,我们将继续两者兼得。

Kimi K3 is a watershed moment because frontier open-weight models are now real. Many hypotheses will be tested on where risks of open-weight models truly land — I suspect it'll be narrower than many expect, and many risks of AI will still be proliferated by closed and "safer" APIs. The evaluation of these risks will evolve in time with an acceleration of AI's integration in our economy. We cannot get one without the other, and we will continue to get both.

因此,2025 年是开放模型开始被更认真对待的一年——尤其是当中国以如此明显的领先优势跃居前列时——因为人们意识到这不会是一个单极世界,不会只由美国封闭的 AI 实验室决定轨迹。2026 年则是之前讨论的、由真正前沿的开放权重模型带来的潜在风险和加速成为现实的一年。

With this, 2025 was when open models started to be taken more seriously — especially when China leaped ahead with such a clear lead — as people realized that it would not be a unipolar world, with only American, closed AI labs determining the trajectory. 2026 is when those previously discussed, potential risks and accelerations due to truly frontier, open-weight models landed.

回应中有很大一部分是针对第四条要点,该条将开放模型的必然结果比作 AI 共产主义,我认为这一说法没有抓住要点。具体而言,在未加解释的情况下使用“共产主义”一词,引发了大部分反弹。

A large portion of the response was for the fourth bullet, which compared the inevitable outcome of open models to AI communism, which I think missed the mark. Specifically, the use of the word communism without explanation caused much of the blowback.

互动版:针对本篇提问 →

接着读 · Read next