{"@context":"https://schema.org","@type":"NewsArticle","generatedAt":"2026-07-28T06:20:51.496Z","headline":"Cursor 的 Agent Swarm 用廉价模型处理大部分编码，前沿模型负责规划","description":"Cursor 在 Cursor 3 中推出 Agent Swarm 产品，将智能体分为 planner（使用 GPT-5.5、Opus 4.8 等前沿模型）和 worker（使用更便宜的 Composer 2.5）。","url":"https://www.aioga.com/news/cms1y7rbi00ekro9f4jv5qkju/","mainEntityOfPage":"https://www.aioga.com/news/cms1y7rbi00ekro9f4jv5qkju/","datePublished":"2026-07-26T15:09:11.000Z","dateModified":"2026-07-26T15:09:11.000Z","inLanguage":"zh-CN","publisher":{"@type":"NewsMediaOrganization","name":"Aioga","url":"https://www.aioga.com"},"citation":["https://the-decoder.com/cursors-agent-swarm-suggests-cheaper-models-can-handle-most-coding-when-frontier-models-plan-the-work","https://aihot.virxact.com/items/cms1y7rbi00ekro9f4jv5qkju"],"canonicalUrl":"https://www.aioga.com/news/cms1y7rbi00ekro9f4jv5qkju/","directAnswer":{"@type":"Answer","text":"Cursor 在 Cursor 3 中将智能体拆分为 planner 与 worker：前沿模型负责递归拆解目标，较便宜的 Composer 2.5 执行具体任务。公司称新系统在重建 SQLite 的 Rust 测试中全部通过。","url":"https://www.aioga.com/news/cms1y7rbi00ekro9f4jv5qkju/","dateCreated":"2026-07-26T15:09:11.000Z","author":{"@type":"Organization","@id":"https://www.aioga.com/authors/aioga-editorial/#editorial-team","name":"Aioga Editorial Team","url":"https://www.aioga.com/authors/aioga-editorial/"}},"evidence":[{"@type":"CreativeWork","name":"the-decoder.com source article","url":"https://the-decoder.com/cursors-agent-swarm-suggests-cheaper-models-can-handle-most-coding-when-frontier-models-plan-the-work","datePublished":"2026-07-26T15:09:11.000Z","provider":{"@type":"Organization","name":"the-decoder.com","url":"https://the-decoder.com/cursors-agent-swarm-suggests-cheaper-models-can-handle-most-coding-when-frontier-models-plan-the-work"}},{"@type":"CreativeWork","name":"AIHot archive record","url":"https://aihot.virxact.com/items/cms1y7rbi00ekro9f4jv5qkju","datePublished":"2026-07-26T15:09:11.000Z","provider":{"@type":"Organization","name":"AIHot","url":"https://aihot.virxact.com/items/cms1y7rbi00ekro9f4jv5qkju"}}],"aggregationSource":"The Decoder：AI News（RSS）","originalPublisher":{"name":"the-decoder.com","url":"https://the-decoder.com/cursors-agent-swarm-suggests-cheaper-models-can-handle-most-coding-when-frontier-models-plan-the-work"},"article":{"id":"cms1y7rbi00ekro9f4jv5qkju","slug":"cms1y7rbi00ekro9f4jv5qkju","url":"https://www.aioga.com/news/cms1y7rbi00ekro9f4jv5qkju/","title":"Cursor 的 Agent Swarm 用廉价模型处理大部分编码，前沿模型负责规划","title_en":"Cursor's agent swarm suggests cheaper models can handle most coding when frontier models plan the work","summary":"Cursor 在 Cursor 3 中推出 Agent Swarm 产品，将智能体分为 planner（使用 GPT-5.5、Opus 4.8 等前沿模型）和 worker（使用更便宜的 Composer 2.5）。","source":"The Decoder：AI News（RSS）","sourceUrl":"https://the-decoder.com/cursors-agent-swarm-suggests-cheaper-models-can-handle-most-coding-when-frontier-models-plan-the-work","aiHotUrl":"https://aihot.virxact.com/items/cms1y7rbi00ekro9f4jv5qkju","publishedAt":"2026-07-26T15:09:11.000Z","category":"产品更新","score":65,"selected":false,"articleBody":["Cursor pitted an upgraded agent swarm against its predecessor by asking both to rebuild SQLite in Rust using only the documentation, with no source code or internet access. Every configuration of the new system reached 100 percent on the test suite, while the old swarm bogged down in its own merge conflicts.","At Cursor, agent fleets have moved from research project to core product. With Cursor 3, developers can run entire fleets of AI agents in parallel：https://the-decoder.com/new-cursor-3-ditches-the-classic-ide-layout-for-an-agent-first-interface-built-around-parallel-ai-fleets/. Anysphere, the company behind Cursor, was recently acquired by Elon Musk's SpaceX for $60 billion：https://the-decoder.com/spacex-bets-60-billion-on-cursor-to-catch-openai-and-anthropic/.","The system divides agents into two roles: planner agents using powerful frontier models recursively break a goal into smaller tasks. Worker agents using faster, cheaper models carry out those tasks. The result is a task tree that adapts as the work progresses.","Cursor says this role split primarily solves a context problem. A lone agent has to traverse the entire tree while keeping both the goal and the current task in mind. That helps explain why agents drift during long jobs. In Cursor's swarm, planners don't write code, and workers don't plan.","An earlier Cursor browser swarm：https://the-decoder.com/cursors-agent-swarm-tackles-one-of-softwares-hardest-problems-and-delivers-a-working-browser/ reached about 1,000 commits per hour on Git. It used worker agents, a judge agent, and an integrator that resolved conflicts. The integrator ended up creating more bottlenecks than it removed.","The new swarm reached 1,000 commits per second, so Cursor built its own version control system because agents working at that rate created failure modes that human teams never encounter.","In what Cursor called \"split-brain design,\" two planners would unknowingly build the same idea in different places and implement it in different ways. Contention was even harder to manage when planners knew about one another and blocked each other with competing edits.","Cursor had agents record decisions in shared design documents. Code tied to a decision linked back to the document through a reference checked at compile time.","When merge conflicts occurred, a neutral agent stepped in and resolved them. Workers flagged bloated files for an outside agent to split into smaller modules. Because agents had learned not to touch core code while working in existing codebases with humans in the loop, Cursor let them break things on purpose. An agent could patch code outside its assigned area, and the compiler would carry the change through the system.","Cursor tested several review approaches. One reviewer received the worker's full transcript, another saw only the output, and a third saw only the codebase. No single perspective caught everything, but uncorrelated perspectives combined for higher reliability.","Cursor also tested a \"field guide,\" a knowledge folder maintained by the agents themselves with a fixed line limit. Every agent received its contents at startup. Since model weights are frozen, it pays to capture surprising findings so later agents can take shortcuts.","Cursor gave the swarm the 835-page SQLite manual and told it to build a Rust implementation. Source code, test suites, the SQLite binary, and internet access were all withheld. The benchmark was sqllogictest, a test suite with millions of SQL queries and known answers. The swarm didn't know it existed.","Four configurations were tested. GPT-5.5 solo, Grok 4.5 solo, Opus 4.8 as planner with Composer 2.5：https://the-decoder.com/cursors-composer-2-5-matches-opus-4-7-and-gpt-5-5-benchmarks-at-a-fraction-of-the-cost/ as worker, and Fable 5：https://the-decoder.com/claude-fable-5-the-first-mythos-model-is-powerful-expensive-and-heavily-filtered/ as planner with Composer 2.5 as worker. The new system beat the old one in every configuration. After four hours, the new runs scored between 73 and 85 percent, while the old runs scored between 11 and 77 percent. Every configuration of the new system later reached 100 percent.","The Grok 4.5 runs showed why the old system fell behind. The old swarm produced 68,000 commits in two hours, about 70 times as many as the new one. Most of that activity was wasted work. The old run accumulated more than 70,000 merge conflicts, while the new run stayed below 1,000 throughout the test.","The most contested file in the old run recorded 7,771 conflicts from 1,173 agents, compared to 47 in the new run. The same split-brain problem appeared in the package structure. The old run broke the project into 54 Rust crates with three separate SQL packages. The new run settled on nine crates early.","In the Fable 5 configuration, the old swarm needed 64,305 lines of engine code, while the new one needed 9,908. In the Opus configuration, the old system produced 19,013 lines and scored 97 percent. The new system reached 100 percent with 4,645 lines.","Total costs ranged from $1,339 for the Opus hybrid to $10,565 for GPT-5.5 running alone. Workers accounted for at least 69 percent of the tokens in every run and usually more than 90 percent. Planner tokens cost more, so the cost split looked different. In the Opus hybrid, the planner produced only a small share of the tokens but accounted for two thirds of the total bill.","The worker model created the largest cost gap. In the GPT-5.5 run, workers alone cost $9,373. In the run using Opus and Composer, the entire worker fleet cost $411 at comparable quality. The difference comes down almost entirely to pricing. Composer 2.5 benchmarks at the level of Opus 4.7 and GPT-5.5 but costs just $0.50 per million input tokens and $2.50 per million output tokens：https://the-decoder.com/cursors-composer-2-5-matches-opus-4-7-and-gpt-5-5-benchmarks-at-a-fraction-of-the-cost/. The model is based on Kimi K2.5, according to Cursor founder Michael Truell：https://the-decoder.com/cursor-quietly-built-its-new-coding-model-on-top-of-chinese-open-source-kimi-k2-5/.","Cursor argues that only a few parts of a large task need the intelligence of a frontier model, including task breakdown and key design decisions. Once a frontier planner resolves the ambiguity, cheaper models can follow its plan, though the hybrid runs showed that planner quality still mattered. The Fable 5 planner used fewer planning tokens than Opus, but its workers needed far more tokens to finish the job. The Fable run cost more overall.","Cursor describes swarms as a kind of probabilistic compiler that translates intent into executable work one step at a time. The company says accurately describing that intent was the main constraint in the experiment. Cursor published the codebase from the Opus solo run as minisqlite on GitHub：https://github.com/cursor/minisqlite.","Runs like these are no longer limited to lab experiments. A prerelease version of Fable 5 handled most of Bun's rewrite from Zig to Rust：https://the-decoder.com/bun-ditches-zig-for-rust-with-help-from-claude-fable-5-writes-over-a-million-lines-of-code-in-11-days/. Sixty-four instances wrote more than a million lines of code in 11 days for about $165,000. Production use still looks different. A study published in late 2025 found that 68 percent of agents used in production completed no more than ten steps：https://the-decoder.com/corporate-ai-agents-use-simple-workflows-with-human-oversight-instead-of-chasing-full-autonomy/ before a human intervened. For 47 percent, the limit was fewer than five.","Stay in the loop on AI. Clear, useful, no fluff.","Follow The Decoder for AI news, background stories and expert analyses.","The Decoder：https://the-decoder.com/"],"articleImages":[{"sourceUrl":"https://the-decoder.com/wp-content/uploads/2026/07/ai-coding-swarms-generated-image-nano-banana-pro.jpg","alt":"Image description","afterParagraph":0,"url":"/media/articles/cms1y7rbi00ekro9f4jv5qkju/8f23ea0d0706db88.jpg"},{"sourceUrl":"https://the-decoder.com/wp-content/uploads/2026/07/cursor-agent-swarm-02-planner-worker-tree.jpg","alt":"Schema \"Decomposing work keeps every agent's context small\": links durchläuft ein einzelner Agent den gesamten Aufgabenbaum, rechts verteilt der Schwarm die Arbeit auf Planner-Knoten und darunterliegende Worker-Knoten.","afterParagraph":3,"url":"/media/articles/cms1y7rbi00ekro9f4jv5qkju/2b6d2ebe07c9182d.jpg"},{"sourceUrl":"https://the-decoder.com/wp-content/uploads/2026/07/cursor-agent-swarm-08-merge-conflicts.jpg","alt":"Liniendiagramm der kumulierten Merge-Konflikte ab dem ersten Konflikt; v1 erreicht nach 120 Minuten fast 70.000 Konflikte mit steigender Kurve, v2 bleibt flach unter tausend.","afterParagraph":13,"url":"/media/articles/cms1y7rbi00ekro9f4jv5qkju/ac1d945561e01a03.jpg"},{"sourceUrl":"https://the-decoder.com/wp-content/uploads/2026/07/cursor-agent-swarm-11-lines-of-code.jpg","alt":"Balkendiagramm der Codezeilen der fertigen Engine; Fable 5 mit Composer 2.5 braucht 64.305 Zeilen unter v1 und 9.908 unter v2, Opus 4.8 mit Composer 2.5 kommt von 19.013 auf 4.645 Zeilen.","afterParagraph":15,"url":"/media/articles/cms1y7rbi00ekro9f4jv5qkju/7dc11ea5b34cae2c.jpg"},{"sourceUrl":"https://the-decoder.com/wp-content/uploads/2026/07/cursor-agent-swarm-01-cost-by-model-mix.jpg","alt":"Gestapeltes Balkendiagramm \"Cost to rebuild SQLite by model mix\" mit Planner- und Worker-Anteilen; Opus 4.8 + Composer 2.5 kostet 1.339 Dollar, Grok 4.5 1.928 Dollar, Fable 5 + Composer 2.5 2.234 Dollar, Opus 4.8 solo 5.153 Dollar, GPT-5.5 10.565 Dollar und Fable 5 solo 20.057 Dollar.","afterParagraph":16,"url":"/media/articles/cms1y7rbi00ekro9f4jv5qkju/72143e662d5baa9c.jpg"},{"sourceUrl":"https://the-decoder.com/wp-content/uploads/2026/07/cursor-agent-swarm-12-tokens-planner-vs-worker.jpg","alt":"Gestapeltes Balkendiagramm der Tokenmenge je Konfiguration, aufgeteilt in Planner und Worker; die Spanne reicht von 2,6 Milliarden Tokens bei Opus 4.8 mit Composer 2.5 bis 14,7 Milliarden bei GPT-5.5.","afterParagraph":18,"url":"/media/articles/cms1y7rbi00ekro9f4jv5qkju/43832ce2b0fe2e37.jpg"}],"mediaStatus":"ok","articleBodyZh":["Cursor 将升级后的代理群与其前身进行对比测试，方法是让两者仅使用文档而不访问源代码或互联网，重建 Rust 中的 SQLite。新系统的每种配置都在测试套件中达到 100% 的通过率，而旧的代理群则因自身的合并冲突而陷入停滞。","在 Cursor，代理队列已从研究项目转变为核心产品。通过 Cursor 3，开发者可以并行运行整支 AI 代理队列：https://the-decoder.com/new-cursor-3-ditches-the-classic-ide-layout-for-an-agent-first-interface-built-around-parallel-ai-fleets/。Cursor 所属公司 Anysphere 最近被埃隆·马斯克的 SpaceX 以 600 亿美元收购：https://the-decoder.com/spacex-bets-60-billion-on-cursor-to-catch-openai-and-anthropic/。","该系统将代理分为两种角色：规划代理使用强大的前沿模型递归地将目标分解为更小的任务；工作代理使用更快、更廉价的模型来执行这些任务。结果是一个随工作进展而自适应的任务树。","Cursor 表示，这种角色划分主要解决了上下文问题。单个代理必须在保持目标和当前任务的同时遍历整个任务树。这也解释了为什么在长时间作业中，代理会发生偏移。在 Cursor 的代理群中，规划者不编写代码，工作者不进行规划。","早期的 Cursor 浏览器代理群：https://the-decoder.com/cursors-agent-swarm-tackles-one-of-softwares-hardest-problems-and-delivers-a-working-browser/ 在 Git 上每小时约能提交 1,000 次。它使用了工作代理、评审代理和解决冲突的整合器。然而，整合器最终制造的瓶颈比它解决的还多。","新的代理群达到了每秒 1,000 次提交，因此 Cursor 自行构建了版本控制系统，因为在这个速度下工作的代理出现的失败模式是人类团队从未遇到的。","在 Cursor 所称的“脑裂设计”中，两个规划者会在不同位置无意识地构建相同的想法，并以不同方式实现。规划者互相知晓对方存在并通过竞争性编辑阻塞对方时，冲突更难管理。","Cursor 让代理在共享设计文档中记录决策。与决策相关的代码通过在编译时检查的引用链接回该文档。","当发生合并冲突时，一个中立的代理会介入并解决它们。工人标记了过大的文件，让外部代理将其拆分为更小的模块。因为代理学会了在有人参与的现有代码库中工作时不触碰核心代码，Cursor 允许它们故意破坏东西。代理可以在其分配区域之外修补代码，编译器会将更改传递到系统中。","Cursor 测试了几种审查方法。一个审查员收到工人的完整记录，另一个只看到输出，第三个只看到代码库。没有单一的视角能捕捉到所有问题，但未相关的多种视角组合提供了更高的可靠性。","Cursor 还测试了“现场指南”，这是一个由代理自身维护的知识文件夹，并有固定的行数限制。每个代理在启动时都会收到其内容。由于模型权重是固定的，捕捉意外发现是有价值的，以便后来的代理可以采取捷径。","Cursor 给群体提供了 835 页的 SQLite 手册，并告诉它构建一个 Rust 实现。源代码、测试套件、SQLite 二进制文件和互联网访问都被禁止。基准测试是 sqllogictest，这是一个拥有数百万 SQL 查询和已知答案的测试套件。群体不知道它的存在。","测试了四种配置。GPT-5.5 单独运行，Grok 4.5 单独运行，Opus 4.8 作为规划者与 Composer 2.5：https://the-decoder.com/cursors-composer-2-5-matches-opus-4-7-and-gpt-5-5-benchmarks-at-a-fraction-of-the-cost/ 作为工人，以及 Fable 5：https://the-decoder.com/claude-fable-5-the-first-mythos-model-is-powerful-expensive-and-heavily-filtered/ 作为规划者，Composer 2.5 作为工人。新系统在每种配置下都击败了旧系统。四小时后，新系统的运行得分在 73% 到 85% 之间，而旧系统的运行得分在 11% 到 77% 之间。新系统的每种配置后来都达到了 100%。","Grok 4.5 的运行显示了旧系统落后的原因。旧群体在两小时内产生了 68,000 次提交，大约是新群体的 70 倍。大部分活动都是无效工作。旧系统的运行积累了超过 70,000 次合并冲突，而新系统的运行在测试期间始终保持在 1,000 次以下。","旧运行中最具争议的文件记录了来自1,173个代理的7,771个冲突，而新运行中只有47个。分裂脑问题在包结构中也出现了。旧运行将项目拆分为54个Rust crate，并有三个独立的SQL包。新运行在早期就确定了九个crate。","在Fable 5配置中，旧的集群需要64,305行引擎代码，而新的只需要9,908行。在Opus配置中，旧系统生成了19,013行代码并得分97%。新系统仅用4,645行代码就达到100%。","总成本从Opus混合的1,339美元到单独运行GPT-5.5的10,565美元不等。每次运行中，工人占至少69%的令牌，通常超过90%。规划器令牌成本更高，因此成本分配看起来有所不同。在Opus混合中，规划器只产生少量令牌，但占总账单的三分之二。","工人模型造成了最大的成本差距。在GPT-5.5运行中，仅工人就花费了9,373美元。在使用Opus和Composer的运行中，整个工人队伍在相当质量下只花费411美元。差异几乎完全归结于定价。根据基准测试，Composer 2.5的性能可与Opus 4.7和GPT-5.5相当，但每百万输入令牌仅需0.50美元，每百万输出令牌仅需2.50美元：https://the-decoder.com/cursors-composer-2-5-matches-opus-4-7-and-gpt-5-5-benchmarks-at-a-fraction-of-the-cost/。据Cursor创始人Michael Truell所述，该模型基于Kimi K2.5：https://the-decoder.com/cursor-quietly-built-its-new-coding-model-on-top-of-chinese-open-source-kimi-k2-5/。","Cursor认为，大任务中只有少部分需要前沿模型的智能，包括任务分解和关键设计决策。一旦前沿规划器解决了歧义，成本更低的模型就可以跟随其计划，尽管混合运行表明规划器的质量仍然重要。Fable 5规划器使用的规划令牌比Opus少，但其工人完成任务所需的令牌远更多。Fable运行的总成本更高。","Cursor 将群体描述为一种概率编译器，它能够将意图一步一步地转化为可执行的工作。公司表示，准确描述这种意图是实验中的主要限制。Cursor 在 GitHub 上以 minisqlite 的形式发布了 Opus 独立运行的代码库：https://github.com/cursor/minisqlite。","像这样的运行不再局限于实验室实验。Fable 5 的预发布版本处理了 Bun 从 Zig 到 Rust 的大部分重写工作：https://the-decoder.com/bun-ditches-zig-for-rust-with-help-from-claude-fable-5-writes-over-a-million-lines-of-code-in-11-days/。64 个实例在 11 天内写了超过一百万行代码，花费约 165,000 美元。生产使用的情况仍然有所不同。一项在 2025 年末发布的研究发现，用于生产的代理中有 68% 在有人干预之前完成的步骤不超过十步：https://the-decoder.com/corporate-ai-agents-use-simple-workflows-with-human-oversight-instead-of-chasing-full-autonomy/。对于 47% 的代理，步骤限制少于五步。","保持对 AI 的关注。清晰、有用，无废话。","关注 The Decoder 获取 AI 新闻、背景故事和专家分析。","The Decoder：https://the-decoder.com/"],"translationStatus":"translated","bodyOrigin":"source-page","editorial":{"summary":"Cursor 在 Cursor 3 中将智能体拆分为 planner 与 worker：前沿模型负责递归拆解目标，较便宜的 Composer 2.5 执行具体任务。公司称新系统在重建 SQLite 的 Rust 测试中全部通过。","background":"该设计针对长任务中的上下文管理问题：单一智能体既要理解总体目标，又要处理局部工作，可能因此偏离方向。新架构让 planner 不写代码、worker 不负责规划，并以任务树随进展调整。","viewpoint":"Aioga 判断，这是一种以角色分工换取成本与执行效率平衡的工程路径，但材料只证明其在特定 SQLite 重建测试中的表现，不能据此推断所有编码任务都适合由廉价模型承担。","implications":"如果类似分工能在更多任务中稳定复现，开发工具的竞争重点可能从单模型能力扩展到任务拆解、并行调度、上下文隔离和版本控制。高并发也可能带来冲突、引用一致性及新型失败管理问题。","nextStep":"值得关注 Cursor 是否披露更多不同类型任务、模型组合与成本数据，以及共享设计文档和编译期引用能否持续降低多智能体协作中的重复实现与编辑冲突。","evidenceRefs":["title","summary","articleBody","source"],"status":"published","aiGenerated":true,"autoApproved":true,"generatedBy":"aioga-editorial:gpt-5.6-sol","reviewedBy":"aioga-editorial-review:gpt-5.6-sol","generatedAt":"2026-07-28T02:14:42.500Z","sourceHash":"fd25166e86aad0fe","review":{"approved":true,"groundedness":95,"clarity":93,"duplicationRisk":12,"blockingIssues":[],"notes":["“全部通过”与来源中“新系统的每种配置均在测试套件中达到 100%”基本一致；若需更精确，可改为“所有配置在测试套件中均达到 100%”。","viewpoint 与 implications 中的推论均通过“Aioga 判断”“如果”“可能”等措辞明确标示为观点或条件性判断，没有冒充已证实事实。","候选内容对单项 SQLite 重建测试的证据边界作了明确限定，避免将结果泛化到所有编码任务。"]},"validation":{"passed":true,"mode":"ai-auto","revisions":0,"checks":["schema","length","source-attribution","low-source-overlap","no-html","independent-ai-review"]}},"tags":["产品更新","The Decoder：AI News（RSS）"],"translations":{"zh-CN":{"title":"Cursor 的 Agent Swarm 用廉价模型处理大部分编码，前沿模型负责规划","summary":"Cursor 在 Cursor 3 中推出 Agent Swarm 产品，将智能体分为 planner（使用 GPT-5.5、Opus 4.8 等前沿模型）和 worker（使用更便宜的 Composer 2.5）。","category":"产品更新","source":"the-decoder.com","aggregationSource":"The Decoder：AI News（RSS）","pageTitle":"Cursor 的 Agent Swarm 用廉价模型处理大部分编码，前沿模型负责规划 - Aioga AI资讯","description":"Cursor 在 Cursor 3 中推出 Agent Swarm 产品，将智能体分为 planner（使用 GPT-5.5、Opus 4.8 等前沿模型）和 worker（使用更便宜的 Composer 2.5）。","url":"https://www.aioga.com/news/cms1y7rbi00ekro9f4jv5qkju/"},"en":{"title":"Cursor's Agent Swarm uses inexpensive models to handle most of the coding, while the cutting-edge models are responsible for planning.","summary":"Cursor launched the Agent Swarm product in Cursor 3, dividing agents into planners (using cutting-edge models like GPT-5.5 and Opus 4.8) and workers (using the more affordable Composer 2.5).","category":"Products","source":"The Decoder：AI News（RSS）","aggregationSource":"The Decoder：AI News（RSS）","pageTitle":"Cursor's Agent Swarm uses inexpensive models to handle most of the coding, while the cutting-edge models are responsible for planning. - Aioga AI News","description":"Cursor launched the Agent Swarm product in Cursor 3, dividing agents into planners (using cutting-edge models like GPT-5.5 and Opus 4.8) and workers (using the more affordable Comp...","url":"https://www.aioga.com/en/news/cms1y7rbi00ekro9f4jv5qkju/","contentTranslated":true,"sourceHash":"3294b606fe52bd83","translatedAt":"2026-07-26T15:43:02.275Z"},"ja":{"title":"Cursor の Agent Swarm は大部分のコーディングを安価なモデルで処理し、最先端モデルが計画を担当する","summary":"Cursor は Cursor 3 で Agent Swarm 製品を発売し、エージェントを planner（GPT-5.5、Opus 4.8 などの最先端モデルを使用）と worker（より安価な Composer 2.5 を使用）に分けました。","category":"製品更新","source":"The Decoder：AI News（RSS）","aggregationSource":"The Decoder：AI News（RSS）","pageTitle":"Cursor の Agent Swarm は大部分のコーディングを安価なモデルで処理し、最先端モデルが計画を担当する - Aioga AIニュース","description":"Cursor は Cursor 3 で Agent Swarm 製品を発売し、エージェントを planner（GPT-5.5、Opus 4.8 などの最先端モデルを使用）と worker（より安価な Composer 2.5 を使用）に分けました。","url":"https://www.aioga.com/ja/news/cms1y7rbi00ekro9f4jv5qkju/","contentTranslated":true,"sourceHash":"3294b606fe52bd83","translatedAt":"2026-07-26T15:43:21.848Z"},"ko":{"title":"Cursor의 Agent Swarm은 저가 모델로 대부분의 코딩을 처리하고, 첨단 모델은 계획을 담당합니다.","summary":"Cursor는 Cursor 3에서 Agent Swarm 제품을 출시하여 에이전트를 planner(GPT-5.5, Opus 4.8 등 최첨단 모델 사용)와 worker(더 저렴한 Composer 2.5 사용)로 나눴습니다.","category":"제품 업데이트","source":"The Decoder：AI News（RSS）","aggregationSource":"The Decoder：AI News（RSS）","pageTitle":"Cursor의 Agent Swarm은 저가 모델로 대부분의 코딩을 처리하고, 첨단 모델은 계획을 담당합니다. - Aioga AI 뉴스","description":"Cursor는 Cursor 3에서 Agent Swarm 제품을 출시하여 에이전트를 planner(GPT-5.5, Opus 4.8 등 최첨단 모델 사용)와 worker(더 저렴한 Composer 2.5 사용)로 나눴습니다.","url":"https://www.aioga.com/ko/news/cms1y7rbi00ekro9f4jv5qkju/","contentTranslated":true,"sourceHash":"3294b606fe52bd83","translatedAt":"2026-07-26T15:44:06.863Z"},"es":{"title":"El enjambre de agentes de Cursor utiliza modelos baratos para manejar la mayor parte del codificado, mientras que los modelos avanzados se encargan de la planificación","summary":"Cursor lanzó el producto Agent Swarm en Cursor 3, dividiendo los agentes en planner (que utiliza modelos de vanguardia como GPT-5.5, Opus 4.8) y worker (que utiliza el más económico Composer 2.5).","category":"Productos","source":"The Decoder：AI News（RSS）","aggregationSource":"The Decoder：AI News（RSS）","pageTitle":"El enjambre de agentes de Cursor utiliza modelos baratos para manejar la mayor parte del codificado, mientras que los modelos avanzados se encargan de la planificación - Aioga Noticias de IA","description":"Cursor lanzó el producto Agent Swarm en Cursor 3, dividiendo los agentes en planner (que utiliza modelos de vanguardia como GPT-5.5, Opus 4.8) y worker (que utiliza el más económic...","url":"https://www.aioga.com/es/news/cms1y7rbi00ekro9f4jv5qkju/","contentTranslated":true,"sourceHash":"3294b606fe52bd83","translatedAt":"2026-07-26T15:44:03.856Z"},"fr":{"title":"L'essaim d'agents de Cursor utilise des modèles bon marché pour traiter la majeure partie du codage, tandis que les modèles de pointe sont responsables de la planification.","summary":"Cursor a lancé le produit Agent Swarm dans Cursor 3, en divisant les agents en planner (utilisant des modèles de pointe comme GPT-5.5, Opus 4.8) et worker (utilisant le moins coûteux Composer 2.5).","category":"Produits","source":"The Decoder：AI News（RSS）","aggregationSource":"The Decoder：AI News（RSS）","pageTitle":"L'essaim d'agents de Cursor utilise des modèles bon marché pour traiter la majeure partie du codage, tandis que les modèles de pointe sont responsables de la planification. - Aioga Actualités IA","description":"Cursor a lancé le produit Agent Swarm dans Cursor 3, en divisant les agents en planner (utilisant des modèles de pointe comme GPT-5.5, Opus 4.8) et worker (utilisant le moins coûte...","url":"https://www.aioga.com/fr/news/cms1y7rbi00ekro9f4jv5qkju/","contentTranslated":true,"sourceHash":"3294b606fe52bd83","translatedAt":"2026-07-26T15:44:48.588Z"},"de":{"title":"Der Agentenschwarm von Cursor verarbeitet den Großteil der Codierung mit günstigen Modellen, während fortschrittliche Modelle für die Planung zuständig sind.","summary":"Cursor hat in Cursor 3 das Produkt Agent Swarm eingeführt und teilt Agenten in Planner (unter Verwendung von Spitzemodellen wie GPT-5.5, Opus 4.8) und Worker (unter Verwendung des günstigeren Composer 2.5) ein.","category":"产品更新","source":"The Decoder：AI News（RSS）","aggregationSource":"The Decoder：AI News（RSS）","pageTitle":"Der Agentenschwarm von Cursor verarbeitet den Großteil der Codierung mit günstigen Modellen, während fortschrittliche Modelle für die Planung zuständig sind. - Aioga KI-News","description":"Cursor hat in Cursor 3 das Produkt Agent Swarm eingeführt und teilt Agenten in Planner (unter Verwendung von Spitzemodellen wie GPT-5.5, Opus 4.8) und Worker (unter Verwendung des...","url":"https://www.aioga.com/de/news/cms1y7rbi00ekro9f4jv5qkju/","contentTranslated":true,"sourceHash":"3294b606fe52bd83","translatedAt":"2026-07-26T15:44:48.629Z"},"pt-BR":{"title":"O Agent Swarm do Cursor lida com a maior parte da codificação usando modelos baratos, enquanto modelos avançados ficam responsáveis pelo planejamento","summary":"A Cursor lançou o produto Agent Swarm no Cursor 3, dividindo os agentes em planner (usando modelos avançados como GPT-5.5, Opus 4.8) e worker (usando o mais barato Composer 2.5).","category":"产品更新","source":"The Decoder：AI News（RSS）","aggregationSource":"The Decoder：AI News（RSS）","pageTitle":"O Agent Swarm do Cursor lida com a maior parte da codificação usando modelos baratos, enquanto modelos avançados ficam responsáveis pelo planejamento - Aioga Notícias de IA","description":"A Cursor lançou o produto Agent Swarm no Cursor 3, dividindo os agentes em planner (usando modelos avançados como GPT-5.5, Opus 4.8) e worker (usando o mais barato Composer 2.5).","url":"https://www.aioga.com/pt-BR/news/cms1y7rbi00ekro9f4jv5qkju/","contentTranslated":true,"sourceHash":"3294b606fe52bd83","translatedAt":"2026-07-26T15:45:30.941Z"},"ru":{"title":"Агентский роевой процессор Cursor использует дешевые модели для обработки большей части кодирования, а передовые модели отвечают за планирование","summary":"Cursor в Cursor 3 выпустил продукт Agent Swarm, разделяя агентов на planner (использующий передовые модели, такие как GPT-5.5, Opus 4.8) и worker (использующий более дешевый Composer 2.5).","category":"产品更新","source":"The Decoder：AI News（RSS）","aggregationSource":"The Decoder：AI News（RSS）","pageTitle":"Агентский роевой процессор Cursor использует дешевые модели для обработки большей части кодирования, а передовые модели отвечают за планирование - Aioga Новости ИИ","description":"Cursor в Cursor 3 выпустил продукт Agent Swarm, разделяя агентов на planner (использующий передовые модели, такие как GPT-5.5, Opus 4.8) и worker (использующий более дешевый Compos...","url":"https://www.aioga.com/ru/news/cms1y7rbi00ekro9f4jv5qkju/","contentTranslated":true,"sourceHash":"3294b606fe52bd83","translatedAt":"2026-07-26T15:45:30.730Z"},"ar":{"title":"عامل Swarm في Cursor يستخدم نماذج رخيصة لمعالجة معظم الترميز، بينما النماذج المتقدمة مسؤولة عن التخطيط","summary":"أطلقت Cursor في Cursor 3 منتج Agent Swarm، حيث تم تقسيم الوكيل إلى planner (باستخدام نماذج متقدمة مثل GPT-5.5 و Opus 4.8) و worker (باستخدام Composer 2.5 الأرخص).","category":"产品更新","source":"The Decoder：AI News（RSS）","aggregationSource":"The Decoder：AI News（RSS）","pageTitle":"عامل Swarm في Cursor يستخدم نماذج رخيصة لمعالجة معظم الترميز، بينما النماذج المتقدمة مسؤولة عن التخطيط - Aioga أخبار الذكاء الاصطناعي","description":"أطلقت Cursor في Cursor 3 منتج Agent Swarm، حيث تم تقسيم الوكيل إلى planner (باستخدام نماذج متقدمة مثل GPT-5.5 و Opus 4.8) و worker (باستخدام Composer 2.5 الأرخص).","url":"https://www.aioga.com/ar/news/cms1y7rbi00ekro9f4jv5qkju/","contentTranslated":true,"sourceHash":"3294b606fe52bd83","translatedAt":"2026-07-26T15:46:16.539Z"},"hi":{"title":"कर्सर का एजेंट स्वर्म सस्ते मॉडल से अधिकांश कोडिंग करता है, अग्रणी मॉडल योजना बनाने के लिए जिम्मेदार होता है","summary":"Cursor ने Cursor 3 में Agent Swarm उत्पाद लॉन्च किया, एजेंट्स को planner (GPT-5.5, Opus 4.8 जैसे उन्नत मॉडलों का उपयोग करते हुए) और worker (सस्ते Composer 2.5 का उपयोग करते हुए) में विभाजित किया।","category":"产品更新","source":"The Decoder：AI News（RSS）","aggregationSource":"The Decoder：AI News（RSS）","pageTitle":"कर्सर का एजेंट स्वर्म सस्ते मॉडल से अधिकांश कोडिंग करता है, अग्रणी मॉडल योजना बनाने के लिए जिम्मेदार होता है - Aioga AI समाचार","description":"Cursor ने Cursor 3 में Agent Swarm उत्पाद लॉन्च किया, एजेंट्स को planner (GPT-5.5, Opus 4.8 जैसे उन्नत मॉडलों का उपयोग करते हुए) और worker (सस्ते Composer 2.5 का उपयोग करते हुए) मे...","url":"https://www.aioga.com/hi/news/cms1y7rbi00ekro9f4jv5qkju/","contentTranslated":true,"sourceHash":"3294b606fe52bd83","translatedAt":"2026-07-26T15:46:20.291Z"},"it":{"title":"L'Agent Swarm di Cursor gestisce la maggior parte della codifica con modelli economici, mentre i modelli all'avanguardia si occupano della pianificazione","summary":"Cursor ha lanciato il prodotto Agent Swarm in Cursor 3, suddividendo gli agenti in planner (che utilizzano modelli all'avanguardia come GPT-5.5, Opus 4.8) e worker (che utilizzano il più economico Composer 2.5).","category":"产品更新","source":"The Decoder：AI News（RSS）","aggregationSource":"The Decoder：AI News（RSS）","pageTitle":"L'Agent Swarm di Cursor gestisce la maggior parte della codifica con modelli economici, mentre i modelli all'avanguardia si occupano della pianificazione - Aioga Notizie IA","description":"Cursor ha lanciato il prodotto Agent Swarm in Cursor 3, suddividendo gli agenti in planner (che utilizzano modelli all'avanguardia come GPT-5.5, Opus 4.8) e worker (che utilizzano...","url":"https://www.aioga.com/it/news/cms1y7rbi00ekro9f4jv5qkju/","contentTranslated":true,"sourceHash":"3294b606fe52bd83","translatedAt":"2026-07-26T15:47:02.857Z"},"nl":{"title":"De Agent Swarm van Cursor verwerkt het grootste deel van de codering met goedkope modellen, terwijl geavanceerde modellen verantwoordelijk zijn voor planning","summary":"Cursor heeft in Cursor 3 het Agent Swarm-product gelanceerd, waarbij agenten worden verdeeld in planners (met gebruik van geavanceerde modellen zoals GPT-5.5 en Opus 4.8) en workers (met gebruik van het goedkopere Composer 2.5).","category":"产品更新","source":"The Decoder：AI News（RSS）","aggregationSource":"The Decoder：AI News（RSS）","pageTitle":"De Agent Swarm van Cursor verwerkt het grootste deel van de codering met goedkope modellen, terwijl geavanceerde modellen verantwoordelijk zijn voor planning - Aioga AI-nieuws","description":"Cursor heeft in Cursor 3 het Agent Swarm-product gelanceerd, waarbij agenten worden verdeeld in planners (met gebruik van geavanceerde modellen zoals GPT-5.5 en Opus 4.8) en worker...","url":"https://www.aioga.com/nl/news/cms1y7rbi00ekro9f4jv5qkju/","contentTranslated":true,"sourceHash":"3294b606fe52bd83","translatedAt":"2026-07-26T15:47:02.105Z"},"tr":{"title":"Cursor'un Agent Swarm'u çoğu kodlamayı ucuz modellerle işliyor, gelişmiş modeller ise planlamadan sorumlu.","summary":"Cursor, Cursor 3'te Agent Swarm ürününü piyasaya sürdü ve ajanları planner (GPT-5.5, Opus 4.8 gibi en son modelleri kullanan) ve worker (daha ucuz olan Composer 2.5'i kullanan) olarak ayırdı.","category":"产品更新","source":"The Decoder：AI News（RSS）","aggregationSource":"The Decoder：AI News（RSS）","pageTitle":"Cursor'un Agent Swarm'u çoğu kodlamayı ucuz modellerle işliyor, gelişmiş modeller ise planlamadan sorumlu. - Aioga AI Haberleri","description":"Cursor, Cursor 3'te Agent Swarm ürününü piyasaya sürdü ve ajanları planner (GPT-5.5, Opus 4.8 gibi en son modelleri kullanan) ve worker (daha ucuz olan Composer 2.5'i kullanan) ola...","url":"https://www.aioga.com/tr/news/cms1y7rbi00ekro9f4jv5qkju/","contentTranslated":true,"sourceHash":"3294b606fe52bd83","translatedAt":"2026-07-26T15:47:47.526Z"},"vi":{"title":"Agent Swarm của Cursor sử dụng mô hình giá rẻ để xử lý phần lớn mã, mô hình tiên tiến chịu trách nhiệm lập kế hoạch","summary":"Cursor đã ra mắt sản phẩm Agent Swarm trong Cursor 3, phân chia các tác nhân thành planner (sử dụng các mô hình tiên tiến như GPT-5.5, Opus 4.8) và worker (sử dụng Composer 2.5 rẻ hơn).","category":"产品更新","source":"The Decoder：AI News（RSS）","aggregationSource":"The Decoder：AI News（RSS）","pageTitle":"Agent Swarm của Cursor sử dụng mô hình giá rẻ để xử lý phần lớn mã, mô hình tiên tiến chịu trách nhiệm lập kế hoạch - Tin tức AI Aioga","description":"Cursor đã ra mắt sản phẩm Agent Swarm trong Cursor 3, phân chia các tác nhân thành planner (sử dụng các mô hình tiên tiến như GPT-5.5, Opus 4.8) và worker (sử dụng Composer 2.5 rẻ...","url":"https://www.aioga.com/vi/news/cms1y7rbi00ekro9f4jv5qkju/","contentTranslated":true,"sourceHash":"3294b606fe52bd83","translatedAt":"2026-07-26T15:47:49.035Z"},"id":{"title":"Agent Swarm dari Cursor menggunakan model murah untuk menangani sebagian besar pengkodean, sementara model canggih bertanggung jawab untuk perencanaan","summary":"Cursor meluncurkan produk Agent Swarm di Cursor 3, membagi agen menjadi planner (menggunakan model mutakhir seperti GPT-5.5, Opus 4.8) dan worker (menggunakan Composer 2.5 yang lebih murah).","category":"产品更新","source":"The Decoder：AI News（RSS）","aggregationSource":"The Decoder：AI News（RSS）","pageTitle":"Agent Swarm dari Cursor menggunakan model murah untuk menangani sebagian besar pengkodean, sementara model canggih bertanggung jawab untuk perencanaan - Berita AI Aioga","description":"Cursor meluncurkan produk Agent Swarm di Cursor 3, membagi agen menjadi planner (menggunakan model mutakhir seperti GPT-5.5, Opus 4.8) dan worker (menggunakan Composer 2.5 yang leb...","url":"https://www.aioga.com/id/news/cms1y7rbi00ekro9f4jv5qkju/","contentTranslated":true,"sourceHash":"3294b606fe52bd83","translatedAt":"2026-07-26T15:48:33.107Z"},"th":{"title":"Agent Swarm ของ Cursor ใช้โมเดลราคาถูกในการจัดการการเข้ารหัสส่วนใหญ่ โมเดลขั้นสูงรับผิดชอบการวางแผน","summary":"Cursor เปิดตัวผลิตภัณฑ์ Agent Swarm ใน Cursor 3 โดยแบ่งเอเจนต์ออกเป็น planner (ใช้โมเดลล้ำสมัย เช่น GPT-5.5, Opus 4.8) และ worker (ใช้ Composer 2.5 ที่ราคาถูกกว่า)","category":"产品更新","source":"The Decoder：AI News（RSS）","aggregationSource":"The Decoder：AI News（RSS）","pageTitle":"Agent Swarm ของ Cursor ใช้โมเดลราคาถูกในการจัดการการเข้ารหัสส่วนใหญ่ โมเดลขั้นสูงรับผิดชอบการวางแผน - ข่าว AI Aioga","description":"Cursor เปิดตัวผลิตภัณฑ์ Agent Swarm ใน Cursor 3 โดยแบ่งเอเจนต์ออกเป็น planner (ใช้โมเดลล้ำสมัย เช่น GPT-5.5, Opus 4.8) และ worker (ใช้ Composer 2.5 ที่ราคาถูกกว่า)","url":"https://www.aioga.com/th/news/cms1y7rbi00ekro9f4jv5qkju/","contentTranslated":true,"sourceHash":"3294b606fe52bd83","translatedAt":"2026-07-26T15:48:40.448Z"},"pl":{"title":"Agent Swarm kursora używa tanich modeli do przetwarzania większości kodowania, zaś nowoczesne modele odpowiadają za planowanie","summary":"Cursor wprowadził w Cursor 3 produkt Agent Swarm, dzieląc agentów na plannerów (korzystających z najnowszych modeli, takich jak GPT-5.5, Opus 4.8) i workerów (korzystających z tańszego Composer 2.5).","category":"产品更新","source":"The Decoder：AI News（RSS）","aggregationSource":"The Decoder：AI News（RSS）","pageTitle":"Agent Swarm kursora używa tanich modeli do przetwarzania większości kodowania, zaś nowoczesne modele odpowiadają za planowanie - Aioga Wiadomości AI","description":"Cursor wprowadził w Cursor 3 produkt Agent Swarm, dzieląc agentów na plannerów (korzystających z najnowszych modeli, takich jak GPT-5.5, Opus 4.8) i workerów (korzystających z tańs...","url":"https://www.aioga.com/pl/news/cms1y7rbi00ekro9f4jv5qkju/","contentTranslated":true,"sourceHash":"3294b606fe52bd83","translatedAt":"2026-07-26T15:49:26.816Z"}}}}