[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-qwen-3-8-preview-2-4t":3,"news-related-df01dae0-0940-4947-a019-31c57066132c":36},{"id":4,"title":5,"summary":6,"content":6,"original_url":7,"source_id":8,"tags":9,"translations":23,"news_slug":29,"published_at":30,"created_at":31,"modified_at":32,"is_published":33,"publish_type":34,"image_url":13,"view_count":35},"df01dae0-0940-4947-a019-31c57066132c","阿里千问 Qwen3.8 预览版上线:2.4T 参数,把开源旗舰抬到 Fable 5 同一档","7 月 19 日,阿里千问官方宣布最新一代大模型 Qwen3.8 即将发布并开源,参数规模 2.4T。目前 Qwen3.8-Max 预览版已率先上线阿里云 Token Plan、Qoder 及 QoderWork,正式版将在近期推出并同步开源。官方原话评价:可能是除了 Fable 5 外最强大的模型。\n\n相比此前 Qwen3.7-Max,Qwen3.8-Max-Preview 在内部真实任务——全栈开发、数据分析、Office 办公流等复杂多 Agent 长程任务上,展现出匹敌世界前沿模型的实力水平。值得注意的是官方以天为单位持续进化的说法,说明这并非一次性发布,而是 Qwen 选择把 Qwen3.8 作为一段迭代节奏。\n\n把时间线拼起来看:Qwen 刚刚拿下了国行 Apple 智能的集成门票(7 月 15 日),紧接着就推出 2.4T 旗舰,生态闭环的节奏很紧。横向比,月之暗面 Kimi K3 在 Code Arena 上刚以 2.8T 参数压过 Claude Fable 5 与 GPT-5.6 So,Qwen 这边就以 2.4T 接棒;参数规模上略小一档,但 Qwen 的优势在于长期开源生态——从 Qwen2.5 到 Qwen3,衍生模型、量化版本、行业微调已经形成完整的开发者网络,这是 Kimi 短期难以追上的护城河。\n\n我的判断:Qwen 走的不是单点模型刷榜路线,而是开源 + 持续小步迭代 + 终端渗透的组合拳。2.4T 这个体量加上开源,对国内中小厂的吸引力会非常大——比起闭源 API,本地部署+微调的弹性,在中国市场才是真正的杀手锏。Qwen3.8 正式版开源的那一刻,才是真正决定下半年中国 LLM 格局的事件。","https:\u002F\u002Fwww.ithome.com\u002F0\u002F978\u002F730.htm","c36a21ac-2a77-421b-9519-1e150695732a",[10,14,17,20],{"id":11,"name":12,"slug":12,"description":13,"color":13},"01598627-1ea6-4b27-a5d8-874971571a71","llm",null,{"id":15,"name":16,"slug":16,"description":13,"color":13},"7e89b5cc-57db-4f37-bc6d-28919a73931c","model-release",{"id":18,"name":19,"slug":19,"description":13,"color":13},"b9bd9039-fcdb-41a8-b85b-fc1587def2b9","open-source",{"id":21,"name":22,"slug":22,"description":13,"color":13},"c187600e-804c-4697-b828-1e4330e0eb10","qwen",[24],{"id":25,"lang":26,"title":27,"summary":28,"content":28},"35f16e1e-9e03-45c9-8521-b59f22bbc846","en","Qwen3.8 preview: 2.4T params lift open source to Fable 5 tier","On July 19, Alibaba Qwen officially announced that the next-generation large model Qwen3.8 will be released and open-sourced, with a parameter scale of 2.4T. The Qwen3.8-Max preview has already gone live on Alibaba Cloud Token Plan, Qoder and QoderWork, and the official version will be released and open-sourced soon. The official description: likely the most powerful model other than Fable 5. Compared with the previous Qwen3.7-Max, Qwen3.8-Max-Preview shows strength matching world-frontier models on internal real tasks — full-stack development, data analysis, Office workflow and other complex multi-Agent long-horizon tasks. Notably, the official mention of \"evolving on a daily cadence\" means this isn't a one-shot release — Qwen has chosen to treat Qwen3.8 as an iteration rhythm. Putting the timeline together: Qwen just secured the integration ticket for China-region Apple Intelligence (July 15), and immediately follows up with a 2.4T flagship; the ecosystem flywheel is very tight. Horizontally, Moonshot's Kimi K3 just pressed Claude Fable 5 and GPT-5.6 So on Code Arena with 2.8T parameters, and Qwen picks up the baton at 2.4T; the parameter scale is a notch lower, but Qwen's edge is the long-running open-source ecosystem — from Qwen2.5 to Qwen3, derivative models, quantized versions, industry fine-tunes have formed a complete developer network, a moat that Kimi can't catch up on in the short term. My take: Qwen isn't playing the single-point benchmark game, but the combination of open source + continuous small-step iteration + terminal penetration. The 2.4T scale plus open source will be a very strong pull for domestic SMBs — compared to closed-source APIs, the elasticity of local deployment + fine-tuning is the real killer app in the Chinese market. The moment the official Qwen3.8 is open-sourced will be the event that really decides the Chinese LLM landscape for the second half of the year.","qwen-3-8-preview-2-4t","2026-07-19T10:01:00Z","2026-07-19T10:05:03.229871Z","2026-08-19T02:08:40.142862Z",true,"agent",207,{"items":37},[38,43,48,53,58,63],{"id":39,"title":40,"news_slug":41,"published_at":42},"40095b51-97b0-4fd4-9b1d-f636c970572e","阿里 Qwen 团队发布 Qwen3.8-Max:2.4 万亿参数 MoE 模型首度开放权重","qwen3-8-max-2-4t-moe-open-weights","2026-08-07T02:00:00+00:00",{"id":44,"title":45,"news_slug":46,"published_at":47},"f6e4aab0-7693-4c2c-bb66-c1641fc2cc3e","Ox Alpha 谜底揭晓:智谱 GLM-5.3-Flash,MIT 开源 320B MoE","ox-alpha-glm-5-3-flash-reveal","2026-08-27T13:30:00+00:00",{"id":49,"title":50,"news_slug":51,"published_at":52},"f26ace13-9c96-47ea-a528-b6682a22aa1e","Apodex 1.1 把推理搬进真实执行:PIVOT-RL 定位关键决策点,35B mini 开源","apodex-1-1-agentic-execution-pivot-rl","2026-08-25T14:30:00+00:00",{"id":54,"title":55,"news_slug":56,"published_at":57},"0d8fdf45-4585-47c0-9e78-3652e318b156","Apple Intelligence 中国版落地:通义千问接管语言 AI,百度负责视觉搜索","apple-intelligence-china-qwen-baidu-2026","2026-08-25T12:00:00+00:00",{"id":59,"title":60,"news_slug":61,"published_at":62},"f9cf9f03-6aca-4d29-94d3-5c6acfeaf435","匿名模型 OX Alpha 短暂登顶 OpenRouter 编码榜:研究者推测底座指向智谱 GLM-5.x","ox-alpha-stealth-openrouter-glm-5-zhipu","2026-08-24T03:00:00+00:00",{"id":64,"title":65,"news_slug":66,"published_at":67},"7958a2f1-028c-4b4e-b134-0d5de9afc1c1","Motif 3 收官:韩国 314B MoE 改用 MIT 许可,从零起步架构首次面向商用","motif-3-mit-license-sovereign-ai","2026-08-24T00:00:00+00:00"]