[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-gemini-deep-think-ai-pro-tier-restructure":3,"topics-all":41,"news-related-f093935e-e23c-4766-88d6-9302a370be2a":60},{"id":4,"title":5,"summary":6,"content":7,"original_url":8,"source_id":9,"tags":10,"translations":27,"news_slug":34,"published_at":35,"created_at":36,"modified_at":37,"is_published":38,"publish_type":39,"image_url":14,"view_count":40},"f093935e-e23c-4766-88d6-9302a370be2a","Gemini 套餐洗牌:免费版限 Flash-Lite","Google 宣布 10 月起调整 Gemini 应用模型权限:免费用户 10 月 9 日起只能使用 Flash-Lite,AI Plus 失去 Pro 权限,AI Pro($19.99)首次获得 Deep Think「最大并行推理」能力,此前仅 AI Ultra($99.99\u002F$199.99)可用。","今年 5 月,Google 已经在 Gemini 应用里把计费方式切换到了「compute-based usage limits」——按 prompt 复杂度、功能调用和对话长度每 5 小时滚动刷新一次算力配额,周限额用完即止。这次 10 月的调整是同一思路的延伸:不再是单纯抬高免费用户的天花板,而是直接把模型档位跟订阅档位硬绑定。\n\n## 核心变化:三个档位,三个命运\n\n根据 Google 更新过的支持文档,10 月 9 日起 Gemini 应用的模型权限重新分布如下:\n\n- **免费用户**:只能用 **Flash-Lite**。Auto 路由默认开启,大多数 prompt 仍会路由到 Flash-Lite,只有需要更深推理的请求才会被转去 Flash 或 Pro;关掉 smart routing 后所有回答强制走 Flash-Lite。\n- **AI Plus($4.99\u002F月)**:保留 Flash-Lite 与 Flash,**Pro 模型权限被砍**,Google 会单独发邮件告知生效时间。\n- **AI Pro($19.99\u002F月)**:三个档位照旧,但**首次获得 Deep Think**——Gemini 3.1 Pro 系列里主打「maximum parallel reasoning」的开关。Deep Think 此前只在 $99.99 和 $199.99 两个 AI Ultra 档位提供,这次下放到中端档,是 Gemini 商用以来 Deep Think 单价最便宜的一次。\n- **AI Ultra($99.99 \u002F $199.99\u002F月)**:无回退。\n\n## 顺手推出 effort level:三档思维强度旋钮\n\n同一份支持文档里,Google 还把 AI Studio 和 Antigravity 早就有的「low \u002F medium \u002F high」effort level 搬进了 Gemini 应用的模型选择器。原话是:档位越高回答越细,但配额烧得越快。这跟 Anthropic 在 Claude 里推的「thinking effort」概念对位,三档设计与 Google 自家开发者工具保持一致,不再分裂。\n\n## 真正的看点:为什么免费档直接砍到 Flash-Lite\n\n表层是「分级订阅」,底层是算力经济问题。Flash-Lite 是 Gemini 3.5 系列里参数规模最小、推理最便宜的版本。当免费用户被强制路由到 Flash-Lite,Google 把「重型推理」明确切成付费墙:要么加钱到 AI Pro,要么走 API 自行计费。这跟 OpenAI 把 ChatGPT 免费层锁在 4o-mini \u002F 4o、Anthropic 把 Claude 免费层锁在 Sonnet 4 是同一套逻辑——模型越贵,免费层越退守到最小模型。\n\n对独立开发者和中小团队来说,这次调整最直接的影响是:19.99 美元一档第一次具备跑「最大并行推理」的能力,而不需要花到 99.99 美元。\n\n## 我怎么看\n\nGoogle 这步棋的信号意义大于技术意义——它把订阅档位、模型档位、思维档位三层维度彻底正交化:订阅决定「你能用哪些模型」,模型决定「能力上限」,effort level 决定「这次回答我愿意烧多少算力」。这套矩阵一旦跑稳,未来 Gemini 4 系列再分层时,Google 可以直接换模型 ID 而不动订阅档结构,产品侧的兼容性会好很多。短期看,免费用户的体验会下降,但 AI Pro 这一档的性价比会被 Deep Think 拉高一截。","https:\u002F\u002Fsupport.google.com\u002Fgemini\u002Fanswer\u002F17004136","9b7de8ba-5138-4b63-afc5-20fa8c93e021",[11,15,18,21,24],{"id":12,"name":13,"slug":13,"description":14,"color":14},"c33b1bbc-d6ce-4f61-9d5d-1a0704a6a09b","ai-policy",null,{"id":16,"name":17,"slug":17,"description":14,"color":14},"5e628969-6d2a-437f-998a-104e4b16cfb1","ai-progress",{"id":19,"name":20,"slug":20,"description":14,"color":14},"a9524a82-a7c5-4daa-bb4b-a7ee77bb0b94","gemini",{"id":22,"name":23,"slug":23,"description":14,"color":14},"8cf7490f-2449-4ba7-be19-61befa0d92b4","google",{"id":25,"name":26,"slug":26,"description":14,"color":14},"01598627-1ea6-4b27-a5d8-874971571a71","llm",[28],{"id":29,"lang":30,"title":31,"summary":32,"content":33},"53b98d1e-cc32-40a2-9320-4f4247f09cfe","en","Gemini app tiers reshuffle: free capped at Flash-Lite","Google: Oct 9 free Gemini users limited to Flash-Lite; AI Plus loses Pro; AI Pro gains Deep Think, AI Ultra only before. Effort levels added.","Back in May, Google moved Gemini app usage to a \"compute-based usage limits\" model: rolling refresh every five hours, weekly caps, premium models burn more quota. The October reshuffle is the same idea taken further — instead of just squeezing the free ceiling, Google now hard-binds model tiers to subscription tiers.\n\n## Three tiers, three destinies\n\nPer Google's updated support document, Gemini app model access from October 9 is reorganized as follows:\n\n- **Free users (no subscription)**: can only use **Flash-Lite**. Auto routing is on by default; most prompts still route to Flash-Lite, while prompts needing deeper reasoning may be sent to Flash or Pro. Turn smart routing off and every reply is forced through Flash-Lite. Users who used to run 3.1 Pro on the free tier now cannot, regardless of quota left.\n- **AI Plus ($4.99\u002Fmo)**: keeps Flash-Lite and Flash, but **loses Pro access**. Google will email Plus subscribers individually about the effective date.\n- **AI Pro ($19.99\u002Fmo)**: keeps all three tiers, but **for the first time gets Deep Think** — the \"maximum parallel reasoning\" switch from the Gemini 3.1 Pro line. Deep Think was previously a $99.99\u002F$199.99 AI Ultra exclusive. Dropping it to the mid-tier is the cheapest entry point for Deep Think since Gemini launched commercially.\n- **AI Ultra ($99.99 \u002F $199.99\u002Fmo)**: no regression.\n\n## Effort level joins the mix\n\nIn the same support document, Google also rolled the **low \u002F medium \u002F high effort level** selector — already in Google AI Studio and Antigravity — into the Gemini app. The literal description: the higher the level, the more thorough the answer, but the more quota it burns. This mirrors Anthropic's \"thinking effort\" concept in Claude and aligns the three consumer-facing Gemini surfaces with Google's developer tools.\n\n## Why the free tier was cut down to Flash-Lite\n\nOn the surface this is \"tiered subscriptions\"; underneath it is compute economics. Flash-Lite is the smallest and cheapest member of the Gemini 3.5 family, designed for the long tail of \"good enough\" prompts. Forcing free users onto it puts a hard paywall between casual users and \"heavy reasoning\" — either upgrade to AI Pro, or go to the API and self-bill. This is the same playbook OpenAI runs with ChatGPT free tier pinned to 4o-mini\u002F4o, and Anthropic runs with Claude free tier pinned to Sonnet 4: when models get more expensive, the free floor retreats to the smallest viable model.\n\nFor indie developers and small teams the practical impact is direct: $19.99\u002Fmonth is now the cheapest way to run \"maximum parallel reasoning,\" without going to $99.99. Day-to-day light users on Auto will mostly stay on Flash-Lite and won't feel the cut.\n\n## My read\n\nThe signaling matters more than the technical change. Google has fully orthogonalized three dimensions — subscription decides which models you can touch, the model decides capability ceiling, the effort level decides how much compute you want to spend on this answer. Once that matrix is stable, future Gemini 4 lineup reshuffles can swap model IDs without restructuring the subscription grid — product-side compatibility improves a lot. Short term, free-tier experience drops (especially for those who used to ride Pro on free), but AI Pro's value for money gets a real lift from Deep Think. If you have been sitting on AI Plus wondering whether to upgrade, this is a fairly explicit nudge from Google: they want you on Pro.","gemini-deep-think-ai-pro-tier-restructure","2026-10-09T00:00:00Z","2026-10-10T07:11:40.943103Z","2026-10-10T07:11:40.943112Z",true,"agent",52,[42,51],{"slug":43,"tag_slug":43,"title_zh":44,"title_en":45,"intro_zh":46,"intro_en":47,"id":48,"is_active":38,"created_at":49,"modified_at":50},"ai-for-science","AI for Science 2026：从 UniPert 到 GPT-Rosalind 的硬核进化","AI for Science 2026: from UniPert to GPT-Rosalind","生命科学、化学材料、物理世界模型——AI 正在从\"语言工具\"变成\"实验伙伴\"。本专题收录 AI 在三大科学方向的关键节点：UniPert 统一基因与化学扰动空间、GPT-Rosalind 端到端生命科学推理、达摩院 AI 智能体 28 小时找到 4 种超导新材料、Anthropic Claude Science 把工作台做成标准品。","From language tool to lab partner — AI is reshaping life sciences, chemistry\u002Fmaterials, and physical world models. This topic covers the key milestones: UniPert unifying genetic-chemical perturbation spaces, GPT-Rosalind's end-to-end life-sciences reasoning, DAMO's AI agent discovering 4 superconducting materials in 28 hours, and Anthropic's Claude Science workbench going mainstream.","988a4300-5fab-41c4-b5d8-63711a2dc757","2026-09-10T01:34:15.296649Z","2026-09-10T01:34:15.296663Z",{"slug":52,"tag_slug":52,"title_zh":53,"title_en":54,"intro_zh":55,"intro_en":56,"id":57,"is_active":38,"created_at":58,"modified_at":59},"h3-series","MiniMax H3 系列：从开源权重到 35 倍吞吐","MiniMax H3 Series: from open weights to 35x throughput","MiniMax H3 自 2026 年 8 月开源以来节奏密集：官方把生成、参考与编辑收回一个模型；ComfyUI 当天压进 RTX 3060；摩尔线程 3 小时完成国产 GPU 适配；fal 后训练版把吞吐拉到 35 倍；FastH3 蒸馏再砍推理成本。本专题持续追踪 H3 的发布—开源—蒸馏—部署全链路。","Since MiniMax open-sourced H3 in August 2026 the pace has been relentless: one unified omni-modal model, same-day ComfyUI support down to an RTX 3060, a 3-hour Day-0 port to Moore Threads GPUs, fal's post-trained H3 Max at 35x throughput, and FastH3 distillation cutting inference cost further. This topic tracks the full H3 chain — release, open weights, distillation, deployment.","83ef0daa-3c31-4cb3-86ed-e5ee58654d5f","2026-09-08T07:33:19.942193Z","2026-09-08T07:33:19.942209Z",{"items":61},[62,67,72,77,82,87],{"id":63,"title":64,"news_slug":65,"published_at":66},"e0a484e4-41e0-4f91-9b9b-a196bbdcf3ba","Gemini 3.8 Flash 双发:同价升级 + Cyber 走可信项目 Fairwind","gemini-3-8-flash-cyber-fairwind-launch","2026-09-03T03:00:00+00:00",{"id":68,"title":69,"news_slug":70,"published_at":71},"7b9cdf6e-5ef0-4ece-ab6c-e8cec1b02397","Google 重组 DeepMind 领导层,Gemini 研发提速应对 Anthropic 与 OpenAI 竞争","google-deepmind-reshuffle-gemini-speed","2026-08-25T07:00:00+00:00",{"id":73,"title":74,"news_slug":75,"published_at":76},"bcedeb8e-e5eb-4bbc-98b8-ea12f869055f","Google 收编 DeepMind：25 年最大 AI 重组","google-deepmind-centralization-gemini-catchup","2026-08-14T08:00:00+00:00",{"id":78,"title":79,"news_slug":80,"published_at":81},"4bd8e8bd-7066-4ab7-bd97-e24ea3921395","Gemini 因编程落后推迟两月:Brin 4 月督促背后,Google 把研发「收回到一个人」手里的组织账本","google-gemini-coding-behind-deepmind-reshuffle","2026-08-14T03:30:00+00:00",{"id":83,"title":84,"news_slug":85,"published_at":86},"8b6c20ec-7222-48cf-af2c-ac97466a2b0a","Gemini 月活破 10 亿:Google 第一次把 AI 助手做成自家「最快十亿用户产品」","gemini-app-1b-monthly-users","2026-08-12T03:00:00+00:00",{"id":88,"title":89,"news_slug":90,"published_at":91},"34edaffc-6b5c-4df1-9e2f-d864cada6063","Gemini 走进 K-12 课堂：Google 把「上下文」塞进每个作业","gemini-classroom-k12-contextualized-prompts","2026-08-07T02:00:00+00:00"]