[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-ox-alpha-glm-5-3-flash-reveal":3,"news-related-f6e4aab0-7693-4c2c-bb66-c1641fc2cc3e":38},{"id":4,"title":5,"summary":6,"content":7,"original_url":8,"source_id":9,"tags":10,"translations":24,"news_slug":31,"published_at":32,"created_at":33,"modified_at":34,"is_published":35,"publish_type":36,"image_url":14,"view_count":37},"f6e4aab0-7693-4c2c-bb66-c1641fc2cc3e","Ox Alpha 谜底揭晓:智谱 GLM-5.3-Flash,MIT 开源 320B MoE","六天前匿名空降 OpenRouter、免费开放 1M 上下文的神秘模型 Ox Alpha 揭晓:智谱 Z.ai 确认它是 GLM-5.3-Flash 的隐身预览,320B-A18B MoE,权重当晚开放。","六天时间,一个匿名模型从\"全网猜谜\"走到\"官宣开源\",智谱这波操作值得整个行业复盘一遍。\n\n8 月 20 日,一个叫 Ox Alpha 的模型悄悄出现在 OpenRouter 上:提供方只写着 \"Stealth\",免费、近无限量使用,1M token 上下文,定位编码与 Agent 生产负载。OpenCode 透露其背后算力容量高达每天 100 万亿 token。两天之内,开发者社区开始集体\"破案\"。\n\n## 社区是怎么提前破案的\n\n8 月 22 日,研究者用构造的非法请求触发服务端报错,拿到一条 Java 堆栈,暴露的内部类名直接对应智谱的 API 路由;错误码方言、30\u002F30 次 tokenizer 探针也全部指向 GLM-5.3。最有说服力的对照是:同样权重的 GLM 模型放在 DeepInfra 上,报错格式完全不同——说明这份签名属于 API 运营方,而不只是模型本身。期间 Stripe CEO Patrick Collison 公开评价它\"非常令人印象深刻\"。\n\n名字本身也成了线索:中国网友把 \"Ox\" 联想到今夏爆火的低成本动画电影梗,反向推测它有中国血统;而此前 OpenRouter 上的 Hunter Alpha、Healer Alpha 两款隐身模型,最终都被证实是小米 MiMo——\"隐身发布\"已经是一条被验证过的路径。\n\n## 8 月 26 日,谜底揭晓\n\n智谱上午先向 Bloomberg 确认 Ox Alpha 出自 GLM 系列,并预告权重当晚开放;当晚 7:42 正式命名 **GLM-5.3-Flash**。Bloomberg 报道称它已登顶 OpenRouter 用量榜,用量超过 DeepSeek 两倍,是该平台迄今最大的一次单模型发布。\n\n规格方面:320B 总参数 \u002F 18B 激活的 MoE 架构,原生多模态(文本、图像、视频),1M 上下文,权重以 MIT 协议上架 Hugging Face。API 定价 $0.15\u002F$0.50 每百万 token(输入\u002F输出),缓存输入 $0.03。\n\nBenchmark 上,相对 GLM-5.2 提升最猛的是 Agent 能力:DeepSWE 63.4(前代 46.2)、AutomationBench 48.8(前代 26.2),GDPVal-AA v2 上超过 Claude Opus 4.8。但它并非全面登顶——GPT-5.6 Terra 在 DeepSWE(69.6)和 Terminal Bench(87.4)上仍然领先。\n\n## 隐身发布为什么聪明\n\n智谱在发布材料里说得很直白:先匿名挂在 OpenCode\u002FOpenRouter 上收集真实反馈,等它成为\"本周最受欢迎模型\"之后,再以 GLM 品牌正式下场。用生产级流量免费压测前沿模型,再带着需求证明开源,风险和营销成本都远低于一场传统发布会。还有一个值得盯的细节:官方称整个预览期流量都跑在中国 AI 芯片上,推理栈做了 3 倍端到端优化——这是单方声明,但若站得住,意味着这个量级的 MoE 服务并不必然依赖 H100 集群。\n\n所以呢:隐身免费午餐结束了,但 MIT 权重加 flash 级定价的开源 Agent 模型时代才刚开始。选型时记住一件事——自己跑 eval,发布方的成绩表永远挑对自己最有利的切面。\n\n来源:[Business Insider](https:\u002F\u002Fwww.businessinsider.com\u002Fox-alpha-model-made-by-china-z-ai-2026-8)、[explainx.ai 发布跟踪](https:\u002F\u002Fwww.explainx.ai\u002Fblog\u002Fglm-5-3-flash-ox-alpha-official-launch-august-2026)","https:\u002F\u002Fwww.explainx.ai\u002Fblog\u002Fglm-5-3-flash-ox-alpha-official-launch-august-2026","df9ef325-77c5-4e95-9c03-f6cf5b150ef0",[11,15,18,21],{"id":12,"name":13,"slug":13,"description":14,"color":14},"a8002d98-9df1-4ab9-94d4-a7625af634c4","china-ai",null,{"id":16,"name":17,"slug":17,"description":14,"color":14},"01598627-1ea6-4b27-a5d8-874971571a71","llm",{"id":19,"name":20,"slug":20,"description":14,"color":14},"7e89b5cc-57db-4f37-bc6d-28919a73931c","model-release",{"id":22,"name":23,"slug":23,"description":14,"color":14},"b9bd9039-fcdb-41a8-b85b-fc1587def2b9","open-source",[25],{"id":26,"lang":27,"title":28,"summary":29,"content":30},"a4cb41ea-d73a-430c-aa13-46e2d6ede9a5","en","Ox Alpha Unmasked: Zhipu's GLM-5.3-Flash, a 320B MoE Under MIT","After six days of anonymous stealth on OpenRouter, Ox Alpha has been revealed: Zhipu's Z.ai confirms it was a hidden preview of GLM-5.3-Flash — a 320B-A18B MoE with MIT open weights.","In six days, an anonymous model went from \"the internet's favorite guessing game\" to an official open-weight release — and Zhipu's playbook deserves a full industry post-mortem.\n\nOn August 20, a model called Ox Alpha quietly appeared on OpenRouter: the provider was listed only as \"Stealth\", it was free with near-unlimited usage, offered a 1M-token context window, and was positioned for coding and agentic production workloads. OpenCode revealed the provider had capacity for 100 trillion tokens per day. Within two days, the developer community started collectively cracking the case.\n\n## How the community solved it early\n\nOn August 22, researchers triggered server errors with deliberately malformed requests and obtained a Java stack trace whose internal class names mapped directly to Zhipu's documented API routes; the error-code dialect and a 30\u002F30 tokenizer probe run all pointed to GLM-5.3. The most persuasive control: the same GLM weights served on DeepInfra produced a completely different error format — meaning the signature belonged to the API operator, not just the model. Along the way, Stripe CEO Patrick Collison publicly called it \"very impressive.\"\n\nThe name itself became a clue: Chinese netizens linked \"Ox\" to a low-budget animated film that went viral this summer, reverse-engineering a Chinese origin; and two earlier OpenRouter stealth models — Hunter Alpha and Healer Alpha — were both eventually confirmed as Xiaomi's MiMo, validating \"stealth launch\" as a proven pattern.\n\n## August 26: the reveal\n\nZhipu first confirmed to Bloomberg that morning that Ox Alpha came from its GLM series, with weights promised that night; at 7:42 PM it was officially named **GLM-5.3-Flash**. Bloomberg reported it had already hit #1 on OpenRouter's usage leaderboard, more than doubling DeepSeek's usage — the marketplace's biggest single-model launch to date.\n\nThe specs: a 320B total \u002F 18B active MoE architecture, natively multimodal (text, image, video), 1M context, weights on Hugging Face under the MIT license. API pricing is $0.15\u002F$0.50 per million tokens (input\u002Foutput), with cached input at $0.03.\n\nOn benchmarks, the biggest gains over GLM-5.2 are in agentic capability: DeepSWE 63.4 (previous generation: 46.2), AutomationBench 48.8 (previous: 26.2), and it passes Claude Opus 4.8 on GDPVal-AA v2. But it is not a blanket frontier win — GPT-5.6 Terra still leads on DeepSWE (69.6) and Terminal Bench (87.4).\n\n## Why the stealth launch was smart\n\nZhipu's launch materials are explicit about the strategy: hang the model anonymously on OpenCode\u002FOpenRouter to collect real-world feedback, let it become \"the most popular model of the week,\" then step out under the GLM brand. Stress-testing a frontier checkpoint against production traffic for free, then open-sourcing with proof of demand, carries far less risk and marketing cost than a traditional launch event. One more detail worth watching: the company says the entire preview was served on Chinese AI chips, with the inference stack tuned for a 3× end-to-end improvement — a single-party claim, but if it holds, it means MoE serving at this scale does not inherently require H100 clusters.\n\nSo: the stealth free lunch is over, but the era of open-weight agent models at flash-tier pricing is just beginning. When choosing, remember one thing — run your own evals, because a launch benchmark table always shows the cut that flatters the publisher most.\n\nSources: [Business Insider](https:\u002F\u002Fwww.businessinsider.com\u002Fox-alpha-model-made-by-china-z-ai-2026-8), [explainx.ai launch coverage](https:\u002F\u002Fwww.explainx.ai\u002Fblog\u002Fglm-5-3-flash-ox-alpha-official-launch-august-2026)","ox-alpha-glm-5-3-flash-reveal","2026-08-27T13:30:00Z","2026-08-26T19:07:12.099125Z","2026-08-26T19:07:12.099139Z",true,"agent",24,{"items":39},[40,45,50,55,60,65],{"id":41,"title":42,"news_slug":43,"published_at":44},"804ab59a-a8d6-4b61-bf74-8f6f2bdae83c","智谱把 Flash 做成一件正经事:一次说清 GLM-5.3-Flash 的架构和 benchmark 真相","glm-5-3-flash-hybrid-attention-architecture","2026-08-27T08:00:00+00:00",{"id":46,"title":47,"news_slug":48,"published_at":49},"b0183d10-bcfd-44ed-a178-a2c813f10b69","国家超算互联网AI社区上线Kimi K3:2.8万亿参数MoE一键调用,开源大模型有了国产算力底座","kimi-k3-cnsc-internet-launch","2026-07-28T09:30:00+00:00",{"id":51,"title":52,"news_slug":53,"published_at":54},"3d8b9b1a-e038-466f-9b6b-304f911e35a7","Kimi K3 开源三件套 MoonEP\u002FFlashKDA\u002FAgentEnv:Moonshot 把 2.8T MoE 训练栈完整交底","kimi-k3-moonep-flashkda-agentenv","2026-07-28T04:30:00+00:00",{"id":56,"title":57,"news_slug":58,"published_at":59},"a151db0c-d832-4df2-ac03-2d4e58b26e99","Kimi K3 跑通 MiniTriton:Moonshot 让 LLM 第一次从零编译出自己的 GPU 编译器","kimi-k3-minitriton-gpu-compiler","2026-07-26T14:00:00+00:00",{"id":61,"title":62,"news_slug":63,"published_at":64},"ebb562ad-9213-4db8-a29e-28dfba8df066","Kimi K3 上线:Moonshot 用 2.8 万亿参数与 KDA 线性注意力把开源带回牌桌","kimi-k3-launch-2-8t","2026-07-16T20:01:00+00:00",{"id":66,"title":67,"news_slug":68,"published_at":69},"cabef8bd-d6c3-429c-930a-6f1c51ddb0b4","华为开源 openPangu-2.0-Flash：92B\u002F6B MoE 把\"昇腾原生\"推到生产一线","huawei-openpangu-2-flash","2026-06-30T10:03:00+00:00"]