Rohan Paul@rohanpaul_ai · 3天前50

"The scaling of open source models, I think it's going down a very dangerous path. And if, the path continues, I think we could get to a very dangerous place. " ~ Anthropic CEO Dario Amodei https://x.com/CodeByNZ/status/2071496679451148503/video/1

译Anthropic CEO Dario Amodei 表示，开源模型的发展正走向危险路径。他指出，AI 领域的开源与传统软件不同：只能看到模型权重（open weights）而非源码，许多协作优势不适用。他将开源问题视为“red herring”，评估模型时只关心模型本身优劣，称 Deep Seek 的成功与是否开源无关。他还强调开源并非免费，模型需托管在云端运行推理，且需有人优化推理速度。

Chubby♨️@kimmonismus · 3天前56

The worst-case scenario for the United States is becoming increasingly realistic, and I will briefly explain why. @quxiaoyin raised many valid points, and I agree with her. First of all: -China certainly does not place such strong emphasis on open source because it cares so deeply about humanism, but because it is a strategy to attract many users, gain market share, put pressure on US models, and also because the models are increasingly being trained on Huawei hardware (think of DeepSeek 4), allowing China to host the entire stack domestically. -But the underlying logic is far more important: The United States is still building too few data centers to meet future demand. @ChrisGillett wrote an outstanding analysis on this, which I shared a week ago. In short, based on SemiAnalysis data, demand is greater than what is currently being built in terms of data centers. -Even more importantly, however, the United States lacks sufficient energy and grid capacity. This is a problem that will become much more severe in the near future. China, by contrast, is addressing the issue through a massive expansion of its energy supply. Solar capacity: in 2025 alone, China installed as much solar capacity as the United States did in 10 to 15 years. China is also building 36 nuclear power plants, significantly more than the United States, and is installing them faster. -In addition, China is managing to become more independent through Huawei chips, even though the country still lags far behind NVIDIA. But here, China is betting on quantity rather than quality. In short: China is a real threat in the AI race, and the situation for the United States is becoming increasingly precarious. This is also the main reason why China is to be kept away from SOTA LLMs at all costs, so as not to jeopardize the lead under any circumstances.

译Kim引用quxiaoyin观点指出，中国开源AI并非出于人文主义，而是抢占市场份额、压制美国模型的战略。模型在华为芯片（如DeepSeek 4）上训练和推理优化，使中国掌握模型层与芯片层。同时美国数据中心建设不足，能源和电网容量严重短缺；中国2025年一年新增太阳能装机量相当于美国10–15年，在建36座核电站。华为芯片虽仍落后NVIDIA，但以量取胜。出口管制无法解决问题，美国应投资开源模型、推动中国使用NVIDIA并加速核电建设。

meng shao@shao__meng · 3天前16

逗死我了，美国政府 BAN 了 Llama，是因为太强太危险了吗？当然不是，因为太烂了。。拿出来，丢人！

Chubby♨️@kimmonismus · 3天前50

.@emollick has used data from Artificial Analysis to show how the development of intelligence compares with open source. Two interesting points: -The development is still unquestionably exponential. There is no doubt about that. AI is not just getting better, it is getting better faster and faster. This is a truth that all of us probably already notice when using AI. -The gap to open source remains fairly constant. Chinese models in particular are still about half a year behind closed source. But that also means that Mythos-class models as open-source variants are genuinely realistic toward the end of the year. A very interesting graph.

译基于Artificial Analysis的AA-Briefcase评分（模拟多周复杂咨询任务），@emollick 绘制前沿曲线发现：闭源AI模型发展呈指数级增长且加速，开源模型（尤其中国）仍落后约半年。但乐观预测，年底前可能出现“神话级”开源变体。

Rohan Paul@rohanpaul_ai · 3天前55

👀 This is from an ex-Meta PM.

译一位前Meta产品经理预测，欧美企业将抛弃OpenAI和Anthropic，转而采用中国模型，原因有四：1）可在中国模型上自托管于自有GPU，确保合规且更可控；2）可在模型之上用自有数据进行后训练，构建数据护城河；3）不信任Anthropic会以“安全”为由保留数据并复制其业务（详见Fable及医疗法律案例）；4）需要证明AI投入的ROI。他认为，根本解是可靠的美版开源模型，但目前没有。

Chubby♨️@kimmonismus · 3天前23

This is a relatively old clip of Dario Amodei, but his stance on open source apparently hasn’t changed at all.

译这是 Dario Amodei 的一个较旧的片段，但他对开源的立场显然完全没有改变。

Berryxia.AI@berryxia · 3天前64

Supervision：Roboflow出品的计算机视觉开源工具包，斩获45K GitHub Stars，三周涨5K!。 它把最常见的CV工作流抽象成了可复用的组件：模型无关的推理、各种 annotator（框、掩码、标签、轨迹）、数据集加载转换、跟踪和区域统计等。 无论你用YOLO、RF-DETR还是其他检测模型，拿到detections对象后，几行代码就能完成标注和可视化。 最强的地方在于它极大地降低了重复造轮子的成本。 以前写一个检测+跟踪+统计的Pipeline要写好几百行，现在基本能用Supervision直接搭起来。 社区也贡献了大量高质量的annotator和工具，让整个生态越来越完善。 在当前多模型共存的时代，这种“模型无关 + 高质量可视化 + 数据集工具”的组合，基本成了大多数CV项目的默认依赖。 Github项目地址👇🏻

译Roboflow出品的Supervision工具包已获45K GitHub Stars，三周增长5K。它将常见CV工作流抽象为复用组件，支持模型无关推理、多种annotator（框、掩码、标签、轨迹）、数据集加载转换、跟踪及区域统计。用户只需几行代码即可用YOLO、RF-DETR等模型完成检测标注可视化，大幅降低重复造轮子成本，社区贡献了丰富工具生态。

Ethan Mollick@emollick · 3天前54

Even though I made this graph, it is also kind of wrong. Fable is guardrailed Mythos. If we use the Mythos date

译根据@ArtificialAnlys的AA-Briefcase评估（让AI执行多周咨询任务），@emollick绘制了开放与封闭模型的前沿曲线，显示令人惊讶的快速进步，且开放权重模型与封闭模型之间存在明显差距。

SemiAnalysis@SemiAnalysis_ · 3天前5

Transformer’s Attention mechanism has come a long way. We’d like to thank the researchers and the engineers in the open-source community for continuing to make high-performance AI accessible. Please celebrate with us by sharing this post, tagging more contributors, and sharing anecdotes to complete the open history of Attention! (1/8)🧵

译Transformer 的注意力机制已经走过了漫长的道路。我们要感谢开源社区的研究人员和工程师，他们持续让高性能 AI 变得可及。请与我们一同庆祝，分享此帖，标记更多贡献者，并分享趣闻，共同完善注意力的开放历史！（1/8）🧵

Chubby♨️@kimmonismus · 4天前42

The open internet was the training ground for modern AI. Billions of texts, forum posts, code repositories, Wikipedia articles, scientific papers, blogs, and discussions: a global knowledge space built collectively over decades. This free knowledge has now been transferred into artificial intelligence. And that is precisely why AI must not disappear entirely behind proprietary walls. Closed source may currently be leading when it comes to the most powerful models. But if access to intelligence is increasingly regulated, restricted, and controlled geopolitically, then open source is no longer just a technical ideal. It is a question of power. The open internet made AI possible. Now open source must ensure that AI remains open.

译强调开放互联网是现代AI的训练基础——数十亿文本、论坛帖子、代码仓库、百科、论文、博客等共同构建了全球知识空间，这一自由知识已被转移进AI。因此AI不应完全消失于专有壁垒之后。闭源模型当前虽在最强模型上领先，但如果地缘政治上对智能的获取日益受限制、管控和控制，开源就不再只是技术理想，而是权力问题。针对“开源仍无法真正查看模型内部，算不上真‘自由’”的质疑，作者表示观看视频后反而更坚信开源的立场。

AK@_akhaliq · 4天前46

baidu/Unlimited-OCR is now number 1 model on huggingface

译baidu/Unlimited-OCR 现在在 HuggingFace 上排名第一

Chubby♨️@kimmonismus · 4天前19

"open source is kind of a distraction you still can’t really see what’s happening inside the model, so it’s not truly "free." After watching the clip, I’m even more convinced of open source than I was before.

译"开源有点分散注意力，你仍然无法真正看到模型内部发生了什么，所以它不是真正的‘自由’。" 看完这段视频后，我比以前更确信开源了。

Ethan Mollick@emollick · 4天前53

GLM-5.2 is good but it is not GPT-5.5/Opus 4.8, and even further from Mythos. Yet it is solid & it demonstrates that the open models continue to chase the frontier What is happening is that open weights crossed into GPT-5.2 territory & capabilities at that point are considerable

译GLM-5.2 不错，但它不是 GPT-5.5/Opus 4.8，甚至远不如 Mythos。不过它很扎实，表明开源模型继续追赶前沿。正在发生的是，开源权重进入了 GPT-5.2 的领域，而那一层面的能力是相当可观的。

Yuchen Jin@Yuchenj_UW · 4天前40

GLM-5.2 is the open-source Claude moment. The demand we’re seeing at Databricks is astonishing. The world is going to see massive adoption of oss LLMs. Also, more companies will shift toward post-training their own models on top of oss models and owning the weights.

译GLM-5.2 是开源的 Claude 时刻。 我们在 Databricks 看到的（模型）需求令人震惊。全世界将大规模采用开源大语言模型。 此外，更多公司会转向在开源模型之上后训练自己的模型并拥有权重。

Nathan Lambert@natolambert · 4天前43

With everything going on, it gives me hope that there's such a diversity of companies building open models today. A lot of the story of open models unfolds under the shadow of the biggest frontier models. Lots of unearthed value.

译Nathan Lambert 表示，在最前沿模型的阴影下，开源模型生态的多样性令人充满希望，仍有大量未发掘的价值。@interconnectsai 在 Artifacts 22 中盘点了5-6月发布的30个开源模型，来自 NVIDIA（3个）、Cohere（2个）、智谱、Zyphra（3个）、Poolside、月之暗面、阶跃星辰、Google（3个）、MiniMax、微软等22家公司/机构。

Chubby♨️@kimmonismus · 4天前45

The release of GLM-5.2 is the second DeepSeek moment.

译中国开源权重模型 GLM-5.2 发布，被评价为继 DeepSeek 之后的第二次“DeepSeek 时刻”。有评论指出，其性能已可与 OpenAI 和 Anthropic 当前可用模型媲美。白宫 AI 负责人 David Sacks 就此警告，美国若继续将自身模型置于“炼狱”（指过度监管或限制），世界将转向中国技术，美国公司将在竞赛中落后。该言论呼应了此前 DeepSeek 开源模型的全球影响，凸显中美 AI 开源竞争进入新阶段。

Rohan Paul@rohanpaul_ai · 4天前57

I’m hearing that "Owl Alpha", one of OpenRouter’s fastest-growing agent models, is actually Meituan LongCat-2.0-Preview The reported design is a huge 1.6T-parameter MoE, active 48B. A dynamic active range of roughly 33B to 56B, and natively supports a 1M-token context window. "Owl Alpha" has been quietly trialed on OpenRouter for nearly two months and has already become one of the most used agent models globally. The usage numbers are striking. Captured OpenRouter data shows: - #1 on Hermes Agent - #2 on Claude Code - #3 on OpenClaw - 10.1T monthly tokens - 559B daily tokens - +242% monthly growth That is extraordinary for a model still operating under an anonymous name.

译据X用户Rohan Paul爆料，OpenRouter增长最快的智能体模型"Owl Alpha"实为美团LongCat-2.0-Preview。该模型采用1.6T参数MoE架构，激活参数量48B，动态激活范围33B-56B，原生支持1M token上下文窗口。已在OpenRouter秘密测试近两月，成为全球使用最多的AI智能体模型之一。OpenRouter数据显示其排名：Hermes Agent第1、Claude Code第2、OpenClaw第3；月处理token 10.1T，日token 559B，月增长率242%。

Berryxia.AI@berryxia · 4天前56

Google Research在2024年悄悄开源了一个时间序列模型。 除了做预测的人，没人注意到。这是一个错误。 这个模型叫TimesFM。 论文发在ICML 2024，标题是"一个用于时间序列预测的解码器架构基础模型"。 核心思路直接借鉴语言模型：先在海量数据上预训练，然后用同一个模型预测任何新序列，不需要重新训练。 过去几十年，时间序列预测一直是一个数据集一套模型的模式。 你收集某个问题的数据，选一个模型架构。 在这个数据上训练，验证。如果问题变了，从头来过。 每个数据集都是一个独立项目。 每个场景都是一条独立流水线。 TimesFM改变了这件事，它在大量跨领域、跨频率的时间序列数据上预训练。 训练完成后，面对任何新的时间序列都能直接预测，零样本预测。 2025年9月，Google发布了2.5版本。 参数从500M降到200M，上下文从2048拉到16K。 加了一个30M的分位数预测头，能同时输出点预测和10%到90%的置信区间。 更小的模型。更长的上下文。 更好的结果。这很少见。 实际影响很具体，200M参数跑一张GPU就行。 16K上下文意味着你可以喂五年日数据，模型能抓住年度季节性。 分位数预测头意味着你不只有一个预测值，还有不确定性范围。 Google内部已经在用了。BigQuery ML里用SQL直接调。Google Sheets的Connected Sheets里内置了。Vertex AI提供了Docker端点。 开源版本免费，两行Python。 加载模型，调用forecast。输入numpy数组，输出预测结果。 2026年4月，Google加了通过HuggingFace Transformers和PEFT用LoRA微调的能力。 这意味着你可以用少量领域数据把预训练模型适配到你的具体场景。 时间序列预测不是一个光鲜的领域。没有病毒式传播的演示。没有十亿美元的消费产品。 但每个管理库存、预测需求、监控设备、交易金融工具的企业都依赖它。 TimesFM把这个行业最好的工具变成了pip install就能用的东西。 地址见评论区👇🏻

译Google Research 于2024年开源时序预测基础模型TimesFM（ICML 2024），采用预训练+零样本预测范式。2025年9月发布的2.5版本参数从500M降至200M，上下文窗口扩展至16K，新增30M分位数预测头，可同时输出点预测及10%-90%置信区间。200M参数单GPU可运行，16K上下文支持五年日数据。模型已内置在BigQuery ML、Google Sheets、Vertex AI中，开源版本通过pip install即可使用。2026年4月通过HuggingFace Transformers和PEFT支持LoRA微调，便于领域适配。

Rohan Paul@rohanpaul_ai · 4天前65

Anthropic CEO Dario Amodei on Open-Source AI Models. "I don't think open source works the same way in AI that it has worked in other areas. Primarily because with open source you can see the source code of the model. Here we can't see inside the model, it's often called open weights instead of open source to kind of distinguish that. But a lot of the benefits, which is that many people can work on it and that it's kind of additive, don't quite work in the same way. So I've actually always seen it as a red herring. When I see a new model come out I don't care whether it's open source or not. If we talk about Deep Seek I don't think it mattered that Deep Seek is open source. I think I ask, is it a good model? Is it better than us at the things that matter? That's the only thing that I care about. It actually doesn't matter either way. Because ultimately you have to host it on the cloud. The people who host it on the cloud do inference. These are big models, they're hard to do inference on. When I think about competition I think about which models are good at the tasks that we do. I think open source is actually a red herring. It's not free. You have to run it on inference and someone has to make it fast on inference." --- From 'Alex Kantrowitz' YT channel (full video link in comment)

译Anthropic CEO Dario Amodei 认为，开源在 AI 领域与传统开源不同，属于“红鲱鱼”（干扰因素）。他指出，当前 AI 模型通常只开放权重而非源代码，无法看到内部逻辑，因此传统开源带来的多方协作、累加式改进等优势不成立。他不在乎模型是否开源，只关心其在关键任务上的表现。此外，大模型必须托管在云端进行推理，并非免费，任何用户都需要支付推理成本。该观点出自 Alex Kantrowitz 的 YouTube 访谈。

Rohan Paul@rohanpaul_ai · 5天前69

🇨🇳🇺🇸Chinese AI models are up to 50 times cheaper than their American counterparts on a per-token basis. Especially Qwen, DeepSeek and Kimi, which pressures OpenAI and Anthropic pricing. - From J.P. Morgan report titled "Semiquincententacles: the US grip on global markets at 250" found that The report said Chinese firms accounted for over 45% of all traffic on the AI aggregation platform OpenRouter by April 2026, up from under 2%in late 2024. Some other findings from the report. - Enterprise AI tokens may become commoditized, because many business tasks do not need frontier models and can run on smaller open models. - AI has driven 65%-80% of S&P 500 returns, profits and capex since ChatGPT, creating clear signs of investor overexcitement in semiconductors. - NVIDIA still dominates AI accelerators, but custom chips from Google, Amazon, Microsoft and Meta are gaining because they can cut total cost by 30%-40%. - China is catching up in AI, with better models, rising GPU self-sufficiency and possible chip-scaling workarounds despite export controls. - Taiwan is the US AI system’s weak point, because TSMC supports much of the world’s advanced chip supply while Taiwan is highly exposed to energy and food blockades.

译J.P. Morgan报告显示，中国AI模型每token比美国便宜50倍，Qwen、DeepSeek、Kimi施压OpenAI和Anthropic定价。到2026年4月，中国公司在OpenRouter流量占比将从不足2%升至超45%。报告还指出企业AI token将商品化，多数任务无需前沿模型；AI已驱动标普500回报的65%-80%；NVIDIA仍主导AI加速器，但定制芯片可降总成本30%-40%；中国GPU自给率提升。UBS调查发现，60%监控AI预算的企业已转向更便宜模型，通过模型路由将简单任务分流至Qwen、DeepSeek、MiniMax等开源模型，以应对最高$35K/月账单及团队超配额200%的压力。

OpenRouter@OpenRouter · 5天前53

Four open-weight models have crossed into territory where they are powering real agentic pipelines. New post in our Insights blog about why companies are choosing them in June: https://openrouter.ai/blog/insights/the-open-weight-models-that-matter-june-2026/

译四个开放权重模型已进入能驱动真实智能体管道的领域。 我们的Insights博客新文章，关于为何公司在6月选择它们：https://openrouter.ai/blog/insights/the-open-weight-models-that-matter-june-2026/

swyx 🔜 @aiDotEngineer@swyx · 5天前51

An interesting way to take Noam at his word in regards to always keeping a constant inference budget for any eval reporting - is that open models have a lot more dollar per token mileage than closed model APIs. So anyone launching an open model today or situationally incentivized toward open models should obviously report thinking levels measured by dollar inference on popular inference providers, instead of by number of tokens on the x axis

译swyx引用OpenAI研究员Noam Brown的观点，强调任何评估报告都应保持恒定推理预算。由于开源模型每美元可获得的token量远超闭源API，因此发布开源模型时，应按主流推理提供商的美元成本（而非token数量）来报告思考水平。该观点源自@saranormous与Noam Brown的播客，他们讨论了大规模测试时计算的后果——模型被给予1000万美元预算处理单一任务，并探讨了基准测试失效、计算预算扩展、能力随投入增长及安全等问题。

Nathan Lambert@natolambert · 5天前38

I've been getting a lot more hate than usual as I try to speak my mind about regulatory capture / unintentional attacks on open-source. It's pretty sad, as there are few people in AI that can speak their mind (most companies say they cannot) and I know many people agree with me silently. I also get people saying that you only say that because it supports the outcomes you want, in a weirdly derogatory way. Of course this is true, but I'm choosing to turn down meaningful wealth so I CAN fight for these values, working at non profits to speak my mind. Building a future that is more inclusive, diverse in the application of AI, and fairer for our children. I may not always be right, but it has been clear to me for a while that more openness right now will help way more than supporting the closed causes. I continue to re-visit this and don't think everything should be open like some of the open-source absolutists. I also don't like a lot of my comrades making fun of anthropic, calling the people there evil, etc. Those are not the case. Trying to stay the course!

译AI研究员Nathan Lambert发文称，因公开批评监管俘获（regulatory capture）及无意中对开源发起的攻击，他遭到比以往更多的敌意。他认为业内很少有人能自由发声，许多人私下赞同他的观点。Lambert选择在非营利组织工作、放弃大量财富，以捍卫更开放、包容、公平的AI应用未来。他并非绝对开源主义者，也不认为一切都要开源，同时不满同路人嘲笑Anthropic的行为。他强调当前更多开放性比支持封闭事业更有益。

Ethan Mollick@emollick · 5天前56

The talk around openness & AI needs to distinguish between the vibrant and innovative open source movement that is advancing the state-of-the-art on harnesses & other key areas and open weights frontier models, which are entirely dependent on a the goodwill of a few Chinese firms

译关于开放性与AI的讨论需要区分充满活力和创新的开源运动（它在束缚及其他关键领域推进了技术前沿）与开放权重前沿模型，后者完全依赖于少数中国公司的善意。

Yuchen Jin@Yuchenj_UW · 5天前38

DeepSeek is the GOAT. 🐳 They just published DSpark, a new speculative decoding method that boosts throughput by 51% to 400%. They also open-sourced DeepSpec, the training framework behind it. This is the real open AI.

译DeepSeek 是 GOAT。🐳 他们刚刚发布了 DSpark，一种新的推测解码方法，将吞吐量提升 51% 到 400%。 他们还开源了背后的训练框架 DeepSpec。 这才是真正的开放 AI。

meng shao@shao__meng · 5天前49

突然想到一个问题：美国政府可以限制他们国家的 AI 团队（OpenAI 和 Anthropic）按照政府的要求对外提供模型，范围、条件、节奏都可以限制，甚至一刀切。 但是对几个月后会追上的中国模型，怎么办？他们大部分都是开源的啊！ 难道要搞一个美国版 GFW？再限制美国模型托管团队们不能推理中国模型？

译推文作者提出一个矛盾：美国政府可以要求OpenAI、Anthropic等国内AI团队按政府指令限制模型对外提供的范围、条件和节奏，甚至一刀切。但几个月后中国模型就会追上，且大部分是开源的。届时美国要如何应对？难道要搞一个美国版GFW，再限制美国模型托管团队不能推理中国模型？

Chubby♨️@kimmonismus · 5天前63

Apple wants to buy RAM from China. Everyone now wants to use open-source models from China because of all the drama surrounding Anthropic, OpenAI, and the U.S. government. What comes next: asking whether Huawei will now also sell GPUs to the West for training new models? What a crazy time.

译苹果正游说特朗普政府，希望获准从被列入五角大楼黑名单的中国芯片制造商长鑫存储（CXMT）采购DRAM。内存价格上涨已迫使苹果提高MacBook和iPad售价，导致市值蒸发2630亿美元。苹果称内存价格已“不可持续”，目前其DRAM供应商为美光、三星和SK海力士。与此同时，受Anthropic、OpenAI与美国政府纠纷影响，AI业界正转向使用中国开源模型。下一步市场关注华为是否会向西方提供GPU用于模型训练。

Chubby♨️@kimmonismus · 5天前29

It is absolutely crazy how the last two weeks have changed the entire future. It is unprecedented that access to "frontier" models was cut off,and presumably remains cut off forever. It feels like a watershed moment, as if access to the highest level of human intelligence had been blocked. Open source is the future. Open source is the solution. The last two weeks have powerfully demonstrated this.

译过去两周彻底改变了整个未来，这简直不可思议。 对“前沿”模型的访问被切断，而且很可能永远无法恢复，这是史无前例的。 这感觉像是一个分水岭时刻，仿佛通往人类最高智能的通道已被封锁。 开源是未来。开源是解决方案。过去两周有力地证明了这一点。

AYi@AYi_AInotes · 5天前53

这条机器人推文刷爆了， 但我敢说 90% 转发的人都没意识到， 物理 AI 的门槛已经被砸到脚面了。 不是什么AI做出来的动画演示， 是正儿八经的 SO-ARM101 ACT 控制策略， 直接在浏览器里跑仿真， 连半毛钱硬件都不用买， 以前想碰物理 AI 少说要几千块的机械臂入门， 现在点开网页就能上手调参看效果， 最厉害的是它的门槛下沉速度， 如果我们不用付出成本就能接触最前沿的方案， 这个领域的迭代速度会超出所有人的预期 https://x.com/pham_blnh/status/2070160360309698813/video/1

译物理AI的门槛已被大幅降低。SO-ARM101 ACT控制策略可直接在浏览器中运行仿真，无需购买任何硬件。此前接触物理AI至少需要数千元的机械臂入门设备，现在只需打开网页即可上手调参并观察效果。这种门槛的下沉速度意味着，当人们能以零成本接触前沿方案时，该领域的迭代速度将远超预期。

向阳乔木@vista8 · 5天前23

猜测有一些道理，但采购中转站数据的也不止一家。 也有人说，中转站数据并不算好。 GLM 5.2 各项能力都很均衡，并非集中在coding，可能还有很多别的黑科技？

译针对智谱GLM 5.2的讨论，有观点猜测其训练数据可能包含采购的中转站数据，但该说法存在争议，有人认为中转站数据质量并不算好。GLM 5.2各项能力表现均衡，并非仅集中在coding领域，可能还应用了其他技术。

Rohan Paul@rohanpaul_ai · 5天前74

"the share of tokens used for US models on OpenRouter has collapsed" Bloomberg On OpenRouter, US model token share fell from around 70% to around 30% in a year. A clean warning signal. open-weight Chinese models are becoming capable enough for routine work, cheaper to run, easier to customize, and less dependent on permission from a frontier lab.

译OpenRouter 上美国模型 token 使用份额在一年内从约 70% 降至约 30%。UBS 调查显示，60% 关注 AI 预算的公司正转向更便宜模型和开源中国模型，主因是极端账单：用户月花费高达 3.5 万美元、团队超配额 200%、公司从 5 个内部 AI 工具削减至 2 个。企业采用模型路由策略，将简单任务交给低成本模型，保留高级模型用于复杂推理、代码和长上下文任务。中国开源模型 Qwen、DeepSeek、MiniMax、GLM、Kimi 因可本地运行或通过云目录使用，契合企业成本曲线。

Yuchen Jin@Yuchenj_UW · 6天前32

The biggest baller move Sam could make right now is to open source GPT-5.6 on Huggingface and declare that OpenAI’s original mission has been achieved.

译Sam 现在能做出的最大胆之举就是在 HuggingFace 上开源 GPT-5.6，并宣布 OpenAI 的原始使命已经完成。

Orange AI@oran_ge · 6天前62

最近几个对模型的反直觉的观察 1. GLM 5.2 正在取代 Claude sonnet 和 Opus，成为付费用户最爱的模型 2. DeepSeek v4 Pro 依然是大众里最受欢迎的模型 3. GPT 5.5 虽然很强大， 但几乎没人用 观测的方式的是看 cola 的 token 消耗统计 这也侧面说明 cola 和 codex 用户（GPT5.5）的画像是完全不同的

译推文分享了三个反直觉的模型观察：GLM 5.2 正在取代 Claude Sonnet 和 Opus 成为付费用户最爱；DeepSeek v4 Pro 仍是大众最受欢迎模型；GPT 5.5 虽然强大但几乎无人使用。数据来源为 cola 的 token 消耗统计，侧面说明 cola 和 codex（GPT 5.5 用户）画像完全不同。

Nathan Lambert@natolambert · 6天前42

There's a lot of sloppy thinking around open models. You can ban them and make it impossible for US companies to use them, but this won't stop A) global open model progress B) bad actors using them So what exactly is gained by banning open models, including those from China?

译关于开放模型，有很多草率的想法。你可以禁止它们，让美国公司无法使用，但这不会阻止 A) 全球开放模型的进展 B) 恶意行为者使用它们 那么，禁止开放模型（包括来自中国的）到底能得到什么？

elvis@omarsar0 · 6天前56

Great to see the new GPT-5.6 models finally announced. Sad to see this new release strategy where only a select few get access initially. Not a win for our industry IMO. Open-source AI must win!

译很高兴看到新的GPT-5.6模型终于发布了。 遗憾的是，这种新发布策略只让少数人先行体验。 我认为这对行业并非好事。 开源AI必须胜利！

AYi@AYi_AInotes · 6天前45

来兄弟们看下什么才是普通人最容易复制的信息差生意，而且赚钱最暴利的信息差，真的是藏在你看不起的地方啊 我刚刷到这个视频，就顺手去闲鱼搜了下，结果真的忍不住一声卧槽， 官方免费就能装的Codex， 闲鱼上代装服务¥32块钱一单， 悄无声息卖了五十多单， 更离谱的是标¥788卖正版授权的， 居然还有17个人默默下单， 几乎没啥成本啊，这不比炒A股香吗， 所以新工具刚爆发的窗口期， 最先赚到钱的永远不是技术最好的， 是嗅觉敏锐最会把复杂操作打包成傻瓜服务的那些人啊

译官方免费的Codex工具，在闲鱼上被当作信息差生意。有人提供代装服务，每单32元，已售50多单；还有人标价788元卖所谓正版授权，竟有17人下单。几乎零成本，说明新工具爆发窗口期，赚到钱的不是技术最好的，而是嗅觉敏锐、把复杂操作打包成傻瓜服务的人。

Yuchen Jin@Yuchenj_UW · 6天前30

best case: OSS surpasses Mythos, and the gov stops banning GPT-5.6/Fable. worst case: OSS surpasses Mythos, and then decides to stop being open source.

译最佳情况： 开源超越Mythos，政府不再禁止GPT-5.6/Fable。 最差情况： 开源超越Mythos，然后决定不再开源。

Nathan Lambert@natolambert · 6天前15

Organizing a small AI meetup in Seattle tomorrow afternoon (Fremont/Ballard). Reply / dm if interested in coming.

译明天下午在组织一场小型AI见面会（西雅图Fremont/Ballard区域）。有兴趣请回复或私信。

Berryxia.AI@berryxia · 6天前71

兄弟们，记忆赛道太卷了… 又有一个开源工具给AI coding agent装上了“无限记忆”。 叫Memanto。 它能把你每次和agent的完整工作会话保存下来，用AI自动组织和压缩，然后在下次需要时在90ms内把相关上下文找回来。 支持Claude Code、Cursor、Codex、LangGraph、CrewAI等主流工具。 以前每次新开会话，agent就失忆，你得重新讲一遍项目背景、架构决策、之前踩过的坑。 现在它能记住你上一次做到哪了，直接接力继续干。 实现上没有用传统向量数据库，而是通过AI压缩 + 高效检索来控制成本和速度。 安装也极简，只需要pip install memanto。 这其实是在解决agentic coding里一个很基础但很疼的问题：上下文的持久化和高效复用。 记忆做得好，agent才能真正从“一次性工具”变成“长期协作伙伴”。

译开源工具Memanto为Claude Code、Cursor、Codex、LangGraph、CrewAI等主流AI coding agent提供“无限记忆”能力。它自动保存每次完整工作会话，通过AI压缩和组织，在下一次会话时90ms内检索到相关上下文，解决agent每次新开会话失忆、需重新解释项目背景的问题。实现无需传统向量数据库，安装仅需`pip install memanto`。该项目已在GitHub获1k+ stars，免费开源。

Chubby♨️@kimmonismus · 6天前72

If you look at the documents again closely, it was never the US government's goal to take down Fable 5. It was simply a consequence of the impossibility of doing so for Anthropic. (i am re-reading the documents atm to get a even better picture of where we are heading) The single most load-bearing document in the whole affair is the letter itself, obtained in full by Bloomberg four days after it landed. It is on Commerce Department letterhead, addressed to Anthropic CEO Dario Amodei, and its operative sentence is dry to the point of menace: a license is now required for the "export, reexport, or transfer (in-country)" of Fable 5 and Mythos 5 "to all destinations worldwide and to all 'foreign persons' ... wherever located." It threatens "prompt criminal and civil penalties" and stays in force "until superseded by a subsequent letter." Read that closely, because the whole kill-switch question lives inside it. The license attaches to foreign persons, not to Americans. The letter never orders Anthropic to take the models offline for US citizens. But American export law has a peculiar reach. Under a Cold War provision known as a deemed export, the moment you give a foreign national access to controlled technology, even on US soil, the law treats it as an export to that person's home country. A German engineer logging into Fable 5 from an office in California becomes, on paper, an export to Germany. And because Anthropic cannot sort its users by passport in real time across an API, the only way to comply with a rule aimed at every foreign person on the planet was to pull the plug for everyone. A rule that targeted only foreign users could be obeyed only by taking the model away from all of them. Therefore, the question arises whether the US government may have acted too shortsightedly and is now itself, also with Anthropic, looking for a solution that simultaneously secures its national supremacy.

译美国商务部致函Anthropic，要求对Fable 5与Mythos 5的“出口、再出口或国内转让”实施全球许可证管制，对象涵盖所有外国人士。基于冷战时期“视为出口”条款，外国人在美境内访问受控技术也被视为向国籍国出口。因Anthropic无法通过API实时区分用户国籍，唯一合规方案是向所有用户关闭模型。推主质疑美国政府短视，目前正与Anthropic寻求兼顾国家安全与可用的解决方案。引用认为，能力封锁将使开源模型更具吸引力，尤其利好中国开源，OpenAI与Anthropic将因此受损。