Meta 撼动前沿：开放与闭源

美国开放权重模型回归！ US open weights are back!

这一周，美国开放权重模型的格局发生了一次向**前沿**的跃迁式转变。Meta 以 Muse Spark Open 领跑这一波浪潮，它目前是美国制造的最强开放模型，且领先幅度相当大；而 Reflection 的首个模型 Beam 的加入也同样重要。若没有这两个模型，中国公司构建并发布的开放权重模型中，将有 20 多个优于西方最好的模型。在撰写本文时，Mistral 也发布了一款强劲的模型 Mistral Large 4，指数得分为 38（见下文），但其权重尚未发布。也有传言称，其他机构还将推出几款强劲的美国开放权重模型！

This has been a week where the balance of American open-weight models took a step change back toward the frontier. Meta took the lead in this charge with Muse Spark Open, which is now the strongest American-made open model by quite some margin, but the addition of Reflection's first model Beam is also relevant. Without these two models, there would be 20+ open-weight models built and released by Chinese companies better than the best Western models. At the time of writing this, Mistral has also announced a strong model, Mistral Large 4, with an index score of 38 (see below), but its weights are not yet released. There are rumors of a few more strong American open-weight models coming from other organizations too!

在许多方面，美国与中国最佳模型之间的差距已大致缩小了一半。我们已从“美国最佳开放模型对追求最前沿性能的消费者基本无关紧要”，转变为“它们虽落后但仍具相关性”。

In many ways, the gap between the best American and Chinese models has been reduced roughly by half. We've gone from the best American open models being largely irrelevant to consumers who require the cutting edge of performance, to them being behind but relevant.

这描绘出前沿模型业务**残酷竞争**的本质。Reflection 和 Mistral _公布了_ 其模型得分，而 Meta 在另外两家实验室能够发布各自权重之前，就放出了更聪明模型的权重。所有这些模型都在与数月前发布的中国模型竞争，业界共同在问——自 GLM-5.3 和 Kimi K3 分别于 8 月 18 日和 7 月 16 日发布以来，Kimi 和智谱 AI 都去哪儿了？

This paints the picture of the cut-throat, competitive nature of the frontier model business. Reflection and Mistral _announced_ their model scores and Meta dropped the weights of a smarter model before these other two labs could get their own weights out. All of these models are competing with Chinese models released months ago, where the industry is collectively asking — where have Kimi and Zhipu AI been since the releases of GLM-5.3 and Kimi K3 on Aug. 18th and Jul. 16th, respectively?

这是一个非常微妙的问题，我将在以后深入探讨。今天的文章聚焦于 Meta 的逆转、Muse Spark Open 对生态系统意味着什么，以及西方前沿实验室之间的力量平衡。

That is a very nuanced question, which I'll dig into another time. Today's piece is focused on Meta's turnaround, what Muse Spark Open means for the ecosystem, and the balance of power across Western frontier labs.

为何 Muse Spark Open 应会腾飞——以及为何可能不会 Why Muse Spark Open should take off — and why it may not

在 Interconnects，通过我们的开放模型分析，我们一次又一次地看到：西方开放权重模型比具有相同性能与规模足迹的中国模型获得了惊人的更多采用（尽管中国模型通常先推出）！Muse Spark Open 正进入这一环境，与此同时，开放式 LLM 智能体终于成为现实（约 GLM 5.2 能力及以上）。

One thing we see again and again at Interconnects through our open model analysis is that Western open-weight models get astonishingly more adoption than Chinese models with the same performance-to-size footprint (even though the Chinese models usually come out first)! Muse Spark Open is landing into this environment, at the same time as when open-ended LLM agents are finally real (~GLM 5.2 capabilities and on).

乘着中国模型性能的快速提升，开放生态系统正经历需求加速。推理提供商正在快速扩张，开放后训练经济正在形成，无数企业希望拥有自己的智能。Muse Spark Open 的性能位于 GLM 5.2 之上，接近 DeepSeek V4.1 Flash，落后于 GLM 5.3 或 Kimi K3。它应该完全“足够好”，能够直接融入其中。

Riding off a rapid increase in performance of Chinese models, the open ecosystem has been living through an acceleration in demand. The inference providers are rapidly scaling, the open post-training economy is forming, and countless enterprises are looking to own their intelligence. Muse Spark Open’s performance falls somewhere above GLM 5.2, near DeepSeek V4.1 Flash, and behind GLM 5.3 or Kimi K3. It should be squarely “good enough” to slot right into this.

一些背景信息，来自 Hugging Face 上的 Muse Spark Open 模型卡：

For some background, from the Muse Spark Open model card on Hugging Face:

以下是各种 Muse 模型在 Artificial Analysis 智能指数上的分布。

Here’s the distribution of the various Muse models on the Artificial Analysis Intelligence Index.

包含此内容是为了回溯到原始模型卡片段。据分享，这个开放权重模型经过了额外的安全训练，为此次开放发布做准备。许多观察者认为此次发布可能是一个在安全训练期间被“破坏”的模型，因为它是在 Muse Spark 1.3 之后推出的。

This is included to loop back to the original model card snippet. It was shared that this open-weight model underwent additional safety training to prepare for this open release. Many observers saw this release as potentially a model that was “broken” during the safety training, as it came out after Muse Spark 1.3.

实际上，这更可能是对 Muse Spark 1.2 进行的安全微调，在 Artificial Analysis 上仅下降 1 分以增强安全行为。当 Meta 发布 30B 智能体模型 Muse Glimmer 时，Alex Wang 表示他们正准备以开放权重形式发布 Muse Spark 1.2 模型，这与上述情况相符。

In reality, this was much more likely a safety tune of Muse Spark 1.2, which would only be a 1 point drop on Artificial Analysis to shore up safety behaviors. When Meta released Muse Glimmer, the 30B agentic model, Alex Wang said that they were preparing to release the Muse Spark 1.2 model as open weights, which fits with the above.

如果说有什么会阻碍 Muse Spark Open，那将是他们为其设计的特定许可证。总之，这个自定义许可证几乎包含了过去几年在构建开放模型时我们所见过的（并在此批评过的）所有限制性许可证行为。具体如下：

If anything will hold Muse Spark Open back, it’ll be the specific license they designed for it. In summary, this custom license has pretty much all of the restrictive license behaviors we’ve seen (and criticized here) in the last few years of building with open models. They are as follows:

1. 在产品中使用该模型的用户必须在产品上显示“Built with Muse”。Meta 在 Llama 上这样做过，Kimi 也尝试过，但几乎没有人遵守。

1. Users of the models in products must display “Built with Muse” on their products. Meta did this with Llama, and Kimi has tried it, but almost no one complies.

2. 存在一个收入门槛因素：整个业务中部署该模型每月收入超过 2000 万美元的公司需要与 Meta 签订单独许可证。其他知名的开放模型提供商也在这样做，以便为转售推理建立收入分成协议。

2. There is a revenue-gating factor, where companies that make more than $20M per month across their entire business deploying the model need a separate license with Meta. Other prominent open model providers are doing this too, in order to set up revenue share agreements for re-selling inference.

3. Muse Spark Open 要求你同意可接受使用政策，在真正的“开源软件”圈子里，这使得该模型极其不开源。

3. Muse Spark Open makes you agree to an acceptable use policy, which in the true “open source software” corner of the world makes this model extremely not open-source.

4. 上述所有内容同样适用于模型的输出及其衍生品，因此，能否安全地将该模型用于合成数据存在法律上的模糊性。

4. All of the above applies to outputs from the model as well, and derivatives, so there is legal ambiguity if you can safely use the model for synthetic data.

该许可证**确实**允许将 Muse Spark Open 用于内部研究和应用。

The license _does_ permit using Muse Spark Open for internal research and applications.

我们处于一个奇怪的均衡状态：该许可证复杂且在法庭上未经检验，但许多参与者可能会忽略它。众多开发者将通过各类推理提供商基于此模型进行构建，而 Meta 可能只会追究严重违规行为。

We’re at a weird equilibrium where this license is complex and unproven in court, but many players will likely ignore it. Many, many people will build on this model from various inference providers, and Meta will only likely go after egregious violations.

目前，许可证尚未成为 AI 采纳的决定性因素，但未来可能会改变。这就是为什么我们认为，对于试图竞争并重新获得采纳的实验室来说，为其模型采用复杂、非标准的开放许可证是一种不必要的战略风险。

For now, licenses haven’t been the determining factor of AI adoption, but this could change in the future. This is why we think it’s an unnecessary strategic risk for labs trying to compete and regain adoption to have complex, non-standard open licenses for their models.

Interconnects AI 是一家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.

Meta 是 Tier 1（OpenAI/Anthropic）还是 Tier 2（Gemini/xAI）前沿实验室？ Is Meta a tier 1 (OpenAI/Anthropic) or tier 2 (Gemini/xAI) frontier lab?

就在 Muse Spark Open 发布前几天，Meta 揭晓了其闭源前沿模型 Muse Wonder 和 Muse Wonder Full。1 在传统基准测试上，这些模型非常强大，仅次于 Anthropic 的 Opus 5.5，与 OpenAI 的 GPT-6-Astra 或 Google 的 Gemini-4-Argon 相似。

Just a few days before Muse Spark Open, Meta unveiled their closed frontier models, Muse Wonder and Muse Wonder Full.1 On traditional benchmarks, these are very strong models only behind Anthropic’s Opus 5.5 and similar to OpenAI’s GPT-6-Astra or Google’s Gemini-4-Argon.

这些重大发布距离 Meta 的大规模 AI 转向仅过去一年多一点，该转向以 CEO Mark Zuckerberg 的《个人超级智能》一文为标志，当时受到了很多批评。Meta 是唯一一家同时拥有前沿开源和闭源模型的 _美国_ 公司。阿里巴巴的 Qwen 也遵循了类似策略，但更倾向于开放。

These major releases are coming just over a year out from Meta’s massive AI redirection, characterized by CEO Mark Zuckerberg’s Personal Superintelligence post, which got a lot of criticism at the time. Meta is the only _American_ company with frontier open and closed models. Alibaba’s Qwen has followed a similar strategy, but with a far more open-leaning approach.

Muse Spark Open 与 Wonder 基准测试。点击展开。

Muse Spark Open and Wonder Benchmarks. Tap to expand.

所有这些模型的氛围检查都表现扎实，但预期是它们不会立即挑战 Anthropic 和 OpenAI 最新模型在实际使用中的精细程度。我很兴奋能更多尝试它们——是时候再订阅一个 AI 服务了。

The vibe checks across all of these models have been solid, but the expectations were for them to not immediately challenge the level of refinement of the most recent Anthropic and OpenAI models in real-world usage. I’m excited to try them more — time to sign up for another AI subscription.

我喜欢 Meta 最近的这一系列表现，在流行的 Muse 智能体中推动自己的产品和方向，在有用时取得胜利，比如领先的美国开放权重模型桂冠。无论财务上是否成功，这都是一个独特的视角，符合他们作为企业的历史。

I like this recent run of form from Meta, driving their own products and directions in the popular Muse agent, taking wins like the leading US open-weight crown when useful. Whether or not it’s financially successful, it’s a unique perspective that fits their history as a business.

人们常常感觉，哪家实验室在紧追 Anthropic-OpenAI 领先双雄，就像走马灯一样变化。目前尚不清楚长期来看 Gemini 和 xAI 是否会出局，但这两家在产品提供上比今天的 Meta 要混乱得多。如果你在 The Cheeky Pint 播客上听过扎克伯格和 Alexandr Wang 的谈话，他们显然有一个非常具体的愿景：建立一个高所有权、每位研究员高算力的实验室，我预计这会带来回报。许多实验室因官僚层级和激励错位而消亡。而 Alex Wang 有超过 200 名直接下属——也就是说，这是一个非常扁平的组织，管理层次极少——并且他们发布的模型和产品改进非常一致。

It often feels like there’s a revolving door of which lab is nipping at the heels of the Anthropic-OpenAI leading duo. It’s not clear that in the long run Gemini and xAI are out of it, but those two are much more muddled in their offerings than Meta today. If you listen to Zuck and Alexandr Wang on The Cheeky Pint, they clearly had a very specific vision of a high-ownership, high-compute per researcher lab that I would expect to pay off. Many labs die due to layers of bureaucracy and misaligned incentives. Meanwhile Alex Wang has over 200 direct reports — i.e. it’s a very flat org with minimal management — and they’re releasing very consistent model and product improvements.

我构建模型的许多世界观是，你需要资源、算力、数据和人才（所有这些都依赖于大量资金），但组织它们非常困难。Gemini 有时被描述为一个无法完成任何工作或成长的地方——这夸大了问题，因为许多人喜欢那里，但其中也有几分道理。管理层级很容易阻碍打造出色的模型。

Much of my worldview of building models is that you need the resources, the compute, the data, and the talent (all downstream of large sums of money), but organizing them is very hard. Gemini is sometimes described as a place where it’s impossible to get any work done or grow — which is embellishing the issue, as many people love it there, but there are kernels of truth to it. Layers of management easily get in the way of making a great model.

虽然我今天在赞扬 Meta，但 Llama 的失败常被描述为那种管理方式走得太远，模型决策出于政治而非性能。xAI 则处于自己的轨道上，它可能更接近于人才流动过大或组织性不足，无法长期维持出色的模型。

While I’m praising Meta today, the Llama failure has often been described as that sort of management going too far, with model decisions being made out of politics rather than performance. xAI is in its own lane, where it may be closer to having too much talent turnover or not enough organization to sustain great models for a long period of time.

回顾过去几年，Meta、Gemini 和 xAI 都生产了几乎处于性能前沿的模型，但 OpenAI 和 Anthropic 的与众不同之处在于它们保持前沿地位的持续性。Anthropic 似乎最为沉稳有序，对这些组织应如何运作有计划和愿景，并因此取得了领先。OpenAI 则时冷时热，但设法在部署模型的前沿保持跟进，同时在公开科学突破（从 Navier-Stokes 到 o1 的推理时间缩放）方面处于领先地位。

When we look back at the last few years, all of Meta, Gemini, and xAI have produced models pretty much at the frontier of performance, but what has distinguished OpenAI and Anthropic is their consistency in maintaining frontier status. Anthropic seems the most poised and organized, with a plan and vision for how these organizations should work, and they’ve gained a lead from it. OpenAI has been more hot and cold, but manages to keep up at the frontier of deployed models while being the leader in public scientific breakthroughs (from Navier-Stokes to o1’s inference time scaling).

以下是本文所关注的公司以及代表整体领先开放权重模型的小米 V2.6 Pro 的竞争态势。排名将是 Anthropic > OpenAI > Meta ~ Gemini > xAI ~ 中国开放权重 > 美国开放权重。

Here’s the state of play among the companies that this post focused on, plus Xiaomi V2.6 Pro representing the leading open-weight model overall. The ranking would be Anthropic > OpenAI > Meta ~ Gemini > xAI ~ Chinese Open Weights > U.S. Open Weights.

AI 产业的开放性问题 Open questions for the AI industry

长期以来，我们一直预测前沿模型竞赛将变得过于昂贵，许多参与者难以跟上，如今这一预测或许正在西方模型中成为现实。鉴于所有模型彼此之间如此接近，我们似乎远未达到“一人通吃”的递归自我改进（RSI）局面，但在美国，能够参与竞争的实验室数量正从众多充满惊喜的广泛候选者，缩小到 Anthropic、OpenAI、Meta、Gemini 和 xAI 这五家。如果微软开始发布模型——鉴于他们在 Scaling（规模扩张）科学上发表了强劲成果——他们可能会加入这一名单，因为他们拥有超大规模业务的财务支持，并且可以直接获取 OpenAI 的知识产权用于训练模型。

After a long time of predicting the frontier model game getting too expensive for many to keep up, we may be seeing this play out among Western models. We seem very far from “one person wins” with RSI given how close all the models are to each other, but we’re narrowing down the number of labs who can compete in the US from a much wider net with surprises, to the 5 of Anthropic, OpenAI, Meta, Gemini, and xAI. If Microsoft starts releasing models, as they’re posting strong scaling science results, they could join this list — given their financial backing with a hyper-scale business and direct access to OpenAI’s intellectual property for training models.

我们从 Mistral、Nemotron、Reflection、Arcee 等“第三梯队”选手那里看到的最佳模型，在影响开放或封闭前沿方面正变得越来越困难。这些模型仍然可以获得大量采用，足以开始发展业务，但构建真正前沿模型的成本现在以每个主要模型数十亿美元计，只有巨型公司才能维持。

The best models we’ve seen from the “tier 3” players like Mistral, Nemotron, Reflection, Arcee, etc. have had an increasingly hard time impacting either the open or closed frontiers. These models still can get substantial adoption, enough to start growing a business, but the costs of building true frontier models are now measured in billions per major model and only huge companies can sustain that.

尽管资源规模如此庞大，行业的技术核心在预测模型在实际使用中的成功程度方面，仍然没有很高的命中率。是的，评估是一个很好的代理指标，但为什么像 Anthropic 这样的领先实验室会从 Claude Fable 5 这样的热门模型，到 Claude Opus 5 的彻底失败，然后又凭借 Opus 5.5 回归热门？准确预测这一点可以赚取数十亿美元，实验室肯定在努力，但模型在长尾本质上仍然很大程度上不可知。

Despite this massive scaling in resources, the technical core of the industry still doesn’t have a high hit rate on predicting how successful a model will be in real-world usage. Yes, evaluations are a good proxy, but why do the leading labs like Anthropic go from hit models like Claude Fable 5 to total duds in Claude Opus 5 and then back to hits with Opus 5.5? There are billions of dollars to be made in predicting this accurately, and the labs are surely trying, but the models fundamentally still are largely unknowable in the long-tail.

在竞赛中剩下的所有玩家里，从长期角度看，由于他们丰富的资产负债表，在财务上押注 Meta、Gemini 和微软似乎要容易得多。这是“AI 将需要一段时间，而 OpenAI/Anthropic 并没有秘密武器”的观点。毕竟，这是有史以来收入增长最快的行业——竞争自然会非常激烈。

Of all of the players left in the race, betting on Meta, Gemini, and Microsoft financially seems much easier from a long-term perspective due to their rich balance sheets. This is the “AI will take a while and OpenAI/Anthropic do not have secret sauce” perspective. After all, it’s the fastest revenue growth industry ever — competition will naturally be immense.

目前，Meta 的估值低于 2 万亿美元，而 Anthropic 正以同样的价格接近其轰动性的 IPO。Meta 在相同技术上似乎落后半步，但从长期视角看，它们似乎从几乎相同的起点出发，而且 Meta 拥有极其盈利的核心业务。这两个估值中有一个需要大幅修正。我和朋友们一直在思考一个有趣的思想实验：你预计 Anthropic 在 2027 年 11 月的公开估值会是多少？2028 年呢？

Currently, Meta is valued under $2T and Anthropic is approaching their barnstorming IPO at the same price. Meta seems half a step behind on the same technology, but in the long-term view these seem to be starting from almost the same point and Meta has an extremely profitable core business. One of the two valuations needs to correct massively. A fun thought experiment I’ve been pondering with friends is: What do you expect Anthropic’s public valuation to be in November of 2027? How about 2028?

Meta 以及 OpenAI 和 Anthropic 的“竞争对手”坚持得越久，AI IPO 周围的负面估值压力就越大（更不用说安全风险定价了）。处于绝对前沿可能在财务上动荡不安，但同时也有价值。时间会证明一切——我们知道市场可以长时间保持非理性。

The longer Meta and the “competitors” to OpenAI and Anthropic stick around, the more negative valuation pressure will exist around the AI IPOs (and that’s not to say anything about pricing in safety risks). Being at the absolute frontier could be financially tumultuous, while valuable. Time will tell — we know markets can stay irrational a very long time.

Muse Wonder 和 Muse Wonder Full 大概是不同的预训练运行，其中 Full 是更大的模型，但这一点并不明确，而且这延续了 AI 产业命名糟糕的传统。有可能“full”只是一个更多循环的版本，鉴于最近关于循环架构的所有新闻，但我没有足够的信心将这一假设放入正文。

Muse Wonder and Muse Wonder Full are presumably different pretraining runs, with Full being a larger model, but it’s not clear and this is up there with the AI industry’s traditions of horrible names. There’s a chance that “full” is just a more looped version, given all the recent news on looped architectures, but I’m not confident enough to put that hypothesis in the main text.