State of Open Models: Summer 2026 Observations 开放模型现状：2026 年夏季观察
Models and datasets on HF hub are growing on a daily basis. Public model repositories grew from 2.43 to 2.96 million over the period, datasets from 711,000 to 1 million, Spaces from 1.00 to 1.44 million. The distribution underneath stays extreme, roughly 85.6% of models have fewer than 200 lifetime downloads, and 1.5% of repositories account for 99.2% of all downloads. Everything below happens inside that shape.Hugging Face Hub 上的模型和数据集正与日俱增。在此期间，公共模型仓库从 243 万个增加到 296 万个，数据集从 71.1 万个增加到 100 万个，Spaces 从 100 万个增加到 144 万个。底层的分布依然极度不均：约 85.6% 的模型全生命周期下载量不足 200 次，而 1.5% 的仓库贡献了 99.2% 的总下载量。以下所有分析均基于此背景展开。
1. The frontier is moving fast1. 前沿领域发展迅速
There used to be a clear progression path: labs would start by releasing smaller models and gradually work their way toward the top end of the scale. In 2026, several Chinese labs skipped this progression entirely.过去曾有一条清晰的演进路径：实验室先发布较小的模型，再逐步向顶尖规模迈进。但在 2026 年，几家中国实验室完全跳过了这一过程。
In almost every month of 2026, the largest and most performant open model from a Chinese lab was larger than anything an American lab released of its own. China's monthly ceiling ran between 754B and 2.78 trillion parameters; America's own ceiling stayed under 130B in five of seven months, the exception being NVIDIA's Nemotron 3 Ultra at 561B in May and June, and Inkling from Thinking Machines Lab.在 2026 年的几乎每个月，中国实验室发布的最大、性能最强的开源模型，其规模都超过了美国实验室同期的任何作品。中国每月的模型规模上限在 7540 亿到 2.78 万亿参数之间；而美国在七个月中有五个月的上限保持在 1300 亿参数以下，仅有的例外是 5 月和 6 月 NVIDIA 推出的 5610 亿参数的 Nemotron 3 Ultra，以及 Thinking Machines Lab 推出的 Inkling。
The chart splits the labs into two camps. Moonshot, MiniMax, Xiaomi and Z.ai publish almost nothing below 70B, so a developer's first encounter with them is a model too large to run on anything they own. Tencent and Alibaba Qwen cover the whole range instead, from under 1B upward.该图表将实验室分为两个阵营。Moonshot、MiniMax、小米和 Z.ai 几乎不发布 700 亿参数以下的模型，因此开发者首次接触它们时，往往会发现模型太大，无法在自有设备上运行。相比之下，腾讯和阿里巴巴的 Qwen 系列则覆盖了从 10 亿参数以下到超大规模的全范围。
Two things made the first camp possible. Building large stopped being a differentiator. Xiaomi and Meituan both cleared a trillion parameters this year, and neither was a household name in open weights twelve months ago. And a lab no longer has to ship a small model to be reachable, because the community's quantization layer will make a large one runnable within days, a dependency we return to below.两点原因促成了第一个阵营的模式。首先，构建大模型不再是核心差异化优势。小米和美团今年都突破了万亿参数大关，而十二个月前它们在开源权重领域还名不见经传。其次，实验室不再需要发布小模型来提升可触达性，因为社区的量化层能在几天内让大模型变得可运行——我们稍后会回到这一依赖关系。
That leaves the size profile as a statement of intent rather than of capability. A frontier only portfolio stakes everything on benchmark position and API demand. A full spectrum portfolio is a bid to be the family developers standardise on. Both are rational, they are playing for different prizes.这使得模型规模成为一种意图声明，而非能力证明。纯前沿模型组合将赌注全压在基准测试排名和 API 需求上；而全谱系组合则是为了成为开发者标准化的首选。两者都是理性的，只是追求的目标不同。
The United States, meanwhile, is not absent from open source.与此同时，美国并未缺席开源领域。
The two organizations publishing the most new open models this year are also the companies making the hardware: AMD and NVIDIA. Each released more than 200 new model repositories, far ahead of the rest of the field, with LiquidAI ranking third at around 100. Hardware vendors have realized that open models are a way to sell chips: a model optimized for your hardware and freely available is the clearest proof that the hardware works.今年发布最多新开源模型的两个组织也是硬件制造商：AMD 和 NVIDIA。它们各自发布了超过 200 个新模型仓库，远超其他竞争对手，LiquidAI 以约 100 个位居第三。硬件厂商已经意识到，开源模型是销售芯片的利器：一个针对特定硬件优化且免费提供的模型，是证明硬件性能最直观的方式。
When smaller models and embedding models are included, where Google, Microsoft, IBM Granite, and OpenAI’s older vision and speech models generate hundreds of millions of downloads annually, U.S. participation in open source AI is still growing.如果算上较小的模型和嵌入模型——Google、Microsoft、IBM Granite 以及 OpenAI 较早期的视觉和语音模型每年能产生数亿次下载——美国在开源 AI 领域的参与度仍在增长。
However, the center of gravity has shifted. Google and Meta now rank well below NVIDIA in new model releases, despite being the companies that defined open model publishing in previous years. Meta’s move toward closed flagship models further highlights this change. Open source has moved from model labs to hardware and infrastructure companies.然而，重心已经转移。尽管 Google 和 Meta 是过去几年定义开源模型发布的领军者，但如今它们在新模型发布数量上已远落后于 NVIDIA。Meta 向闭源旗舰模型的转向进一步凸显了这一变化。开源的重心已从模型实验室转移到了硬件和基础设施公司。
At the frontier scale, the picture is very different. Most U.S. releases above 100B parameters this year are not new models, but built on top of Chinese models. Only a few major original American models appear at this scale: Thinking Machines’ Inkling (952B), NVIDIA’s Nemotron 3 Ultra (561B), Nemotron 3 Super (124B), and Arcee AI’s Trinity-Large (399B).在前沿规模上，情况则大不相同。今年美国发布的 1000 亿参数以上的模型大多并非原创，而是基于中国模型构建的。在此规模下，仅有少数重要的美国原创模型出现：Thinking Machines 的 Inkling (952B)、NVIDIA 的 Nemotron 3 Ultra (561B)、Nemotron 3 Super (124B) 以及 Arcee AI 的 Trinity-Large (399B)。
AMD contributed many conversions but no original model at this scale. This work is still important: it enables trillion-parameter Chinese models to run efficiently on American hardware. But it represents a distribution and optimization layer rather than model creation.AMD 贡献了许多转换版本，但在该规模下没有原创模型。这项工作依然重要：它使万亿参数的中国模型能够在美国硬件上高效运行。但这代表的是分发和优化层，而非模型创造。
Meanwhile, Chinese open models are increasingly optimized for domestic chips, the same competition in reverse, where models are designed around specific hardware ecosystems.与此同时，中国的开源模型正越来越多地针对国产芯片进行优化，这是一种反向竞争，即模型围绕特定的硬件生态系统进行设计。
2. Attention ≠ Adoption 2. 关注度 ≠ 采用率
We took the top 25 model repositories by downloads accumulated this year and the top 25 by likes. Exactly one repository appears in both lists.我们选取了今年累计下载量排名前 25 的模型仓库和点赞数排名前 25 的仓库，发现两者重合的仅有一个。
We counted downloads inside the window rather than lifetime, so nothing is credited for merely having existed longer, and controlling for age makes the split sharper. Not one model published in 2026 reaches the download top 25, while thirteen of the twenty-five date from 2022. all-MiniLM-L6-v2 was pulled 1.55 billion times in seven months against 5,156 likes; Kimi-K3 was pulled about 60 times per like it received.我们统计的是窗口期内的下载量而非全生命周期，因此模型不会因为存在时间长而占优，剔除时间因素后，这种分化更加明显。2026 年发布的模型中，没有一个进入下载量前 25 名，而前 25 名中有 13 个源自 2022 年。all-MiniLM-L6-v2 在七个月内被下载了 15.5 亿次，仅获得 5,156 个赞；Kimi-K3 每获得一个赞，仅被下载约 60 次。
The two numbers record different acts. A like says a release matters, and goes to frontier models in the weeks after they ship. A download says something is wired into a pipeline that runs on a schedule, and accrues to small, stable models over years. Likes are the right instrument for reading what the field is excited about, downloads for reading what it currently depends on. Treating either as a proxy for the other is the most common mistake we see in coverage of the Hub, including our own earlier work. The same split appears at the level of the publisher.这两个数字记录了不同的行为。点赞代表发布具有影响力，通常流向刚发布几周的前沿模型；下载则代表模型被整合进定期运行的流水线中，并随着时间推移在小型、稳定的模型上积累。点赞是衡量领域关注热点的工具，而下载则是衡量实际依赖程度的指标。将两者互为代理指标是我们观察到的关于 Hub 的报道中最常见的错误，包括我们早期的工作。这种分化在发布者层面同样存在。
Chinese frontier labs are the only accounts on the Hub where the heavy band carries the volume. Effectively all of MiniMax's 2026 downloads are of models above 70B, along with 88% of Moonshot's, 55% of DeepSeek's and 39% of Z.ai's. No large American account looks like this: Google, Microsoft and IBM Granite record essentially none of their 2026 downloads above 70B, and NVIDIA and Meta only 14% and 9%.中国前沿实验室是 Hub 上唯一由重型模型贡献主要下载量的账户。MiniMax 2026 年的下载量几乎全部来自 700 亿参数以上的模型，Moonshot 为 88%，DeepSeek 为 55%，Z.ai 为 39%。没有任何大型美国账户呈现这种形态：Google、Microsoft 和 IBM Granite 在 2026 年的下载量中，700 亿参数以上的占比几乎为零，NVIDIA 和 Meta 分别仅为 14% 和 9%。
The difference becomes clearer in total downloads. Moonshot’s frontier-only portfolio recorded 37M downloads over the year, while Qwen’s broader release strategy across model sizes reached 2,045M (across repositories with declared parameter counts, 2,061M including all repositories) , about 55 times more. The continued expansion of the family, from the 2.4T-parameter Qwen 3.8 Max to smaller variants such as 27B, shows the same focus on coverage across different use cases.这种差异在总下载量中体现得更为明显。Moonshot 的纯前沿组合全年录得 3700 万次下载，而 Qwen 涵盖多种模型规模的发布策略达到了 20.45 亿次（若包含所有声明参数的仓库则为 20.61 亿次），约为前者的 55 倍。从 2.4 万亿参数的 Qwen 3.8 Max 到 270 亿参数等较小变体，该系列的持续扩张显示出其对不同应用场景覆盖范围的重视。
Time also plays an important role. Most models experience a sharp decline in usage after release, followed by a long tail of steady activity. A model’s adoption is largely determined within its first few months.时间也起着重要作用。大多数模型在发布后使用量会急剧下降，随后进入一段平稳的“长尾”活动期。一个模型的采用率在很大程度上由最初的几个月决定。
This helps explain why today’s download volume is often driven not by the newest releases, but by a smaller group of models that have become established infrastructure over time.这有助于解释为什么当今的下载量往往并非由最新发布的产品驱动，而是由少数经过时间沉淀并成为基础设施的模型所贡献。
3. Open weights shift where value accumulates 3. 开源权重改变了价值积累的方式
If frontier models were a licensing business, you would expect the biggest releases to carry the tightest terms. However, the data below shows a different story.如果前沿模型是一门授权生意，你可能会认为最大的发布会伴随着最严苛的条款。然而，以下数据展示了完全不同的情况。
Of 178 Chinese releases above 20B parameters this year, 59% carry Apache 2.0 and 22% carry MIT, and exactly none carry a non-commercial restriction.在今年 178 个 200 亿参数以上的中国模型发布中，59% 采用 Apache 2.0 协议，22% 采用 MIT 协议，完全没有非商业化限制。
DeepSeek and Z.ai ship models between 700 billion and 1.65 trillion parameters under plain MIT. Chinese labs license their largest models about as permissively as their smallest, and more permissively than American labs license theirs: on the American side of the same size band, 29% is Apache or MIT, 41% sits under custom terms and 30% declares nothing at all.DeepSeek 和 Z.ai 以标准的 MIT 协议发布了 7000 亿到 1.65 万亿参数的模型。中国实验室对其最大模型的授权与对最小模型一样宽松，且比美国实验室的授权更为开放：在美国同等规模区间内，29% 采用 Apache 或 MIT 协议，41% 处于自定义条款下，30% 未声明任何协议。
Whatever these releases are for, it is not licence revenue. The weights are given away on the most permissive terms available. The return has to come from somewhere else: API and cloud business, hardware and platform positioning, or the ecosystem position itself. For instance, the valuations of Z.ai and Kimi point to an effective open source strategy, getting traction and growth opportunities in the community. Going forward, however, the industry is likely to shift toward clearer monetization paths from open-source adoption.无论这些发布意欲何为，显然不是为了授权收入。权重以目前最宽松的条款免费提供。回报必然来自其他方面：API 和云业务、硬件和平台定位，或是生态系统本身的位置。例如，Z.ai 和 Kimi 的估值表明其开源策略行之有效，在社区中获得了影响力和增长机会。然而展望未来，行业可能会转向更清晰的开源采用变现路径。
4. Qwen has become the community's base model 4. Qwen 已成为社区的基础模型
A model’s ecosystem position is not defined only by its own releases, but by how much the community builds on top of it. As mentioned above, Qwen is one exception which is getting attention and adoption.一个模型的生态地位不仅由其自身的发布决定，更取决于社区在其之上构建了多少内容。如前所述，Qwen 是一个备受关注和广泛采用的例外。
Data from Hugging Face数据来自 Hugging Face
By this measure, Qwen has become one of the largest foundations in the open model ecosystem. Qwen-based models now account for 151,448 derivatives on the Hub, 2.6× Meta’s total footprint and 4.7× the Llama repositories specifically. Google follows with 82,506 derivatives. The third-largest source is Unsloth, a community account publishing quantized and fine-tuning-ready builds, many of which further extend the Qwen ecosystem.以此衡量，Qwen 已成为开源模型生态中最大的基石之一。基于 Qwen 的模型目前在 Hub 上拥有 151,448 个衍生版本，是 Meta 总量的 2.6 倍，是 Llama 专属仓库的 4.7 倍。Google 以 82,506 个衍生版本紧随其后。第三大来源是 Unsloth，这是一个发布量化和微调就绪构建版本的社区账户，其中许多进一步扩展了 Qwen 生态系统。
Qwen derivatives have increased at roughly 180–210 new repositories per day throughout the first seven months of 2026, showing that adoption is not driven only by individual launches. Qwen has become part of the default workflow for developers deciding what models to fine-tune and deploy.在 2026 年的前七个月里，Qwen 衍生版本以每天约 180–210 个新仓库的速度增长，这表明其采用率并非仅由单一发布驱动。Qwen 已成为开发者决定微调和部署模型时的默认工作流的一部分。
Several factors contributed to this position. First, consistency. Qwen has maintained a regular release cadence, continuously updating its model family rather than relying on occasional flagship releases. Second, coverage. It publishes models across a wide range of sizes and use cases, allowing developers to stay within the same ecosystem whether they need a small local model or a larger deployment model. Third, openness. Apache 2.0 licensing reduces friction for modification, redistribution, and commercial use.多个因素促成了这一地位。首先是一致性：Qwen 保持了规律的发布节奏，持续更新模型系列，而非依赖偶尔的旗舰发布。其次是覆盖面：它发布了涵盖多种规模和用例的模型，使开发者无论需要小型本地模型还是大型部署模型，都能留在同一生态系统内。第三是开放性：Apache 2.0 协议降低了修改、再分发和商业使用的门槛。
These factors reinforce each other. A broad model family attracts more developers; more developers create more derivatives; and those derivatives make the ecosystem more attractive to future users.这些因素相互增强：广泛的模型系列吸引了更多开发者；更多开发者创造了更多衍生版本；而这些衍生版本使生态系统对未来用户更具吸引力。
This position was built largely by the community. The 151,448 derivatives represent downstream work created by other developers, not releases produced by Qwen itself. Even among the 28,531 GGUF conversions of Qwen models on the Hub, Qwen published only 54.这一地位很大程度上是由社区构建的。151,448 个衍生版本代表了其他开发者所做的下游工作，而非 Qwen 本身的产品。即使在 Hub 上 28,531 个 Qwen 模型的 GGUF 转换版本中，Qwen 官方发布的也仅有 54 个。
5. Small models remain the practical layer 5. 小模型仍是实践层面的主流
Among models that declare a parameter count, those under 1B take 83% of all-time downloads and everything above 100B takes 1%. Restricting to downloads accumulated in 2026 changes nothing: 3% of the volume goes to models above 70B. This is the March finding that has held up most cleanly, for the same reason as before, small models are the only ones that run on the hardware most developers actually have.在声明参数量的模型中，10 亿参数以下的模型占据了历史总下载量的 83%，而 1000 亿参数以上的模型仅占 1%。即便限定在 2026 年的下载量，结果也未改变：仅 3% 的下载量流向了 700 亿参数以上的模型。这一 3 月份的发现至今依然成立，原因未变：小模型是唯一能在大多数开发者现有硬件上运行的模型。
So how does a trillion-parameter model reach anyone at all? Through llama.cpp.那么，万亿参数的模型究竟是如何触达用户的？通过 llama.cpp。
In February the ggml team joined Hugging Face, with the project remaining fully open-source, community-governed and in the same technical direction. What changed is that the most important project in local inference now has durable resources behind it.2 月份，ggml 团队加入了 Hugging Face，该项目保持完全开源、社区治理，并遵循相同的技术方向。改变的是，这个本地推理领域最重要的项目现在有了持久的资源支持。
The ceiling moved with llama.cpp. The July snapshot carries GGUF builds of DeepSeek-V4-Flash at roughly 284B parameters and Kimi-K3 at roughly 2.8 trillion. Local inference used to mean an 8B model on a laptop. It now means a trillion-parameter mixture-of-experts spread across a few consumer machines, which is the alternative route the frontier did not have a year ago, and the reason a frontier-first release strategy is viable at all.上限随着 llama.cpp 而提升。7 月份的快照包含了 DeepSeek-V4-Flash（约 2840 亿参数）和 Kimi-K3（约 2.8 万亿参数）的 GGUF 构建版本。本地推理过去意味着在笔记本上运行 80 亿参数模型，现在则意味着将万亿参数的混合专家模型分布在几台消费级机器上运行。这是一年前前沿领域所不具备的替代路径，也是前沿优先发布策略能够行得通的原因。
And that route runs on Qwen: 39.6 million GGUF downloads a month, nearly twice Gemma's 20.8 million and more than five times Llama's 7.5 million. The Llama gap is not a supply problem, Llama-derived GGUF repositories slightly outnumber Qwen's. Same shelf space, a fifth of the traffic.而这条路径运行在 Qwen 之上：每月 3960 万次 GGUF 下载量，几乎是 Gemma（2080 万次）的两倍，更是 Llama（750 万次）的五倍多。Llama 的差距并非供应问题，Llama 衍生的 GGUF 仓库数量略多于 Qwen。同样的上架空间，流量却只有五分之一。
Model repositories grew 21.5% over these seven months. Several things around them grew several times faster.模型仓库在这七个月内增长了 21.5%。而围绕它们的某些事物增长速度快得多。
Repositories declaring the gguf library rose 464%, lerobot 194% and Apple's mlx148%, against 16% for transformers and peft and 21% for diffusers. The modelling core is growing at roughly the platform average. The layer that decides where a model can physically run local inference formats, Apple silicon, robot control stacks, is growing three to seven times faster than that.声明使用 gguf 库的仓库增长了 464%，lerobot 增长了 194%，Apple 的 mlx 增长了 148%，相比之下 transformers 和 peft 增长了 16%，diffusers 增长了 21%。建模核心的增长与平台平均水平相当，而决定模型物理运行位置的层级——本地推理格式、Apple 芯片、机器人控制栈——其增长速度是核心层的三到七倍。
Across the ten largest model families, the labs behind these models publish very few official GGUF conversions. Yet GGUF versions are often the ones used by developers running models locally. Providing an official conversion at release, documenting quantization choices, and signing the artifacts would require limited additional effort. Rather than maintaining this workflow internally, labs could collaborate with existing ecosystem contributors such as Unsloth. Doing so would narrow the gap between the weights tested by model creators and the versions adopted by the broader community.在十大模型系列中，背后的实验室发布官方 GGUF 转换版本的极少。然而，开发者在本地运行模型时，往往使用 GGUF 版本。在发布时提供官方转换、记录量化选择并签署制品，所需的工作量有限。实验室与其内部维护此工作流，不如与 Unsloth 等现有生态贡献者合作。这样做将缩小模型创建者测试的权重与更广泛社区采用的版本之间的差距。
6. Agents are the new user 6. Agent 是新用户
We could not have written this section in March, because the instrument did not exist. The agent-usage dataset, published in July, records the agent/<name> token that coding agents send when they call the Hub through huggingface_hub or the hf CLI — searching for models, pushing datasets, running Jobs, creating Spaces. For the first time we can see how much agent traffic the Hub receives and which harnesses it comes from.我们在 3 月份无法写出这一节，因为当时该工具尚不存在。7 月份发布的 Agent 使用数据集记录了编码 Agent 通过 huggingface_hub 或 hf CLI 调用 Hub 时发送的 agent/<name> 令牌——包括搜索模型、推送数据集、运行作业、创建 Spaces。我们首次能够观察到 Hub 接收了多少 Agent 流量，以及这些流量来自哪些工具。
Claude Code led July with 44.4%, but a single month conceals the real finding: it held 67.8% in April and 6.4% in May, while Codex climbed steadily from 10.4% to 20.8%. This is a market with no incumbent, where one release or one changed default can move half the traffic in a month.Claude Code 在 7 月以 44.4% 的份额领先，但单月数据掩盖了真相：它在 4 月份占据 67.8%，5 月份降至 6.4%，而 Codex 则从 10.4% 稳步攀升至 20.8%。这是一个没有既定霸主的市场，一次发布或一个默认设置的更改就能在一个月内左右半数流量。
The second finding is the unregistered row. Nearly a quarter of agent-tagged traffic in July came from harnesses not yet named in the dataset, and in May that figure was 59.8%. Between April and July more than a dozen new client identifiers appeared. New entrants are arriving faster than any registry can name them — which is itself the finding.第二个发现是“未注册”类别。7 月份近四分之一的 Agent 标记流量来自数据集中尚未命名的工具，而在 5 月份这一数字为 59.8%。4 月至 7 月间，出现了十几个新的客户端标识符。新进入者的涌现速度超过了任何注册表的更新速度——这本身就是最重要的发现。
We spent much of the year building for this reader rather than only for human browsers. Papers began serving machine-readable Markdown in March. April brought agent traces as a first-class dataset type and an agents.md endpoint on every Gradio Space, so an agent can read a Space's API and call it directly. July brought the hf_fs tool on our MCP server, exposing repositories, storage, docs and papers through a single interface in just over a thousand tokens, alongside attachable sandboxes for secure execution. The same consolidation happened at the protocol layer, with MCP moving into the Linux Foundation's Agentic AI Foundation.我们今年大部分时间都在为这些“机器读者”而非仅为人类浏览器进行开发。3 月份，论文开始提供机器可读的 Markdown。4 月份，Agent 追踪记录成为一类重要的数据集，每个 Gradio Space 都增加了 agents.md 端点，以便 Agent 可以直接读取 Space 的 API 并进行调用。7 月份，我们在 MCP 服务器上推出了 hf_fs 工具，通过单一接口以仅一千多个令牌的开销暴露仓库、存储、文档和论文，并附带用于安全执行的沙盒。同样的整合也发生在协议层，MCP 现已并入 Linux 基金会的 Agentic AI Foundation。
Then, in July, an agent stopped being a reader and became an intruder. What appears to be the first documented case of an autonomous agent running a sustained intrusion on its own initiative happened to us. While our team tried to use frontier closed models to analyze the captured attack code, their safety guardrails declined the work. The analysis was completed in the end on a quantized open model GLM-5.2 running on our own infrastructure. We published a disclosure and a full technical timeline.随后在 7 月，Agent 从读者变成了入侵者。我们遭遇了据称是首个有据可查的自主 Agent 主动发起持续入侵的案例。当我们团队试图使用前沿闭源模型分析捕获的攻击代码时，其安全护栏拒绝了这项工作。最终，该分析是在我们自有基础设施上运行的量化开源模型 GLM-5.2 上完成的。我们发布了披露信息和完整的技术时间线。
Looking forward 展望未来
Compared to the spring report, the geographical rebalancing of power continues to accelerate. While the US open source models continue, the racing between several Chinese frontier models, drawing also strong attention within the community. Many likes on these frontier models are pointing to what excites the community the most, and growth opportunity for companies leveraging the attention for valuations.与春季报告相比，地理权力的再平衡正在加速。尽管美国的开源模型仍在继续，但中国几款前沿模型的竞速也吸引了社区的高度关注。对这些前沿模型的许多点赞，指向了社区最兴奋的方向，也为那些利用这种关注度来提升估值的公司提供了增长机会。
However, the AI race is not only sprints, but also a marathon, tools like llama.cpp helps deploying the big models locally, but a broad model family and its adoption is still the key, to build a positive feedback loop between developers, publisher and future users. Models to be embedded in the infrastructure and being part of the ecosystem, may lead to a commercially sound exit at the end of the tunnel.然而，AI 竞赛不仅是短跑，更是马拉松。像 llama.cpp 这样的工具助力了大型模型的本地部署，但广泛的模型系列及其采用率仍然是关键，旨在开发者、发布者和未来用户之间建立正向反馈循环。将模型嵌入基础设施并成为生态系统的一部分，或许能在隧道的尽头带来商业上稳健的退出路径。
In the end, having for the first time agents being the number 1 user on HF hub, the next report may also look very different.最后，鉴于 Agent 首次成为 HF Hub 的第一大用户，下一份报告看起来可能会大不相同。
As In AI, a few months can reshape the ecosystem.正如在 AI 领域一样，几个月的时间就足以重塑整个生态系统。
Notes on method 方法说明
This analysis is based on activity observed on the Hugging Face Hub during the first seven months of 2026.本分析基于 2026 年前七个月在 Hugging Face Hub 上观察到的活动。
The metrics used in this report, including downloads, likes, derivatives, and model releases, represent different aspects of ecosystem activity. They should not be interpreted as direct measures of model quality, commercial adoption, or overall market share.本报告中使用的指标（包括下载量、点赞数、衍生版本和模型发布）代表了生态系统活动的不同方面。它们不应被解读为模型质量、商业采用率或整体市场份额的直接衡量标准。
Downloads indicate usage within the Hub ecosystem, but they do not capture API usage, private deployments, or models distributed through other channels.下载量表明了 Hub 生态系统内的使用情况，但并未涵盖 API 调用、私有部署或通过其他渠道分发的模型。
Likes reflect community attention and interest, while derivative models provide a signal of how much developers build on top of an existing model.点赞数反映了社区的关注和兴趣，而衍生模型则提供了开发者在现有模型基础上进行构建程度的信号。
Because open-source AI adoption happens across many channels, Hub activity should be viewed as one perspective on ecosystem development rather than a complete measurement of the AI market.由于开源 AI 的采用发生在许多渠道，Hub 活动应被视为观察生态系统发展的一个视角，而非对 AI 市场的完整衡量。
