{"@context":"https://schema.org","@type":"NewsArticle","generatedAt":"2026-07-28T05:20:57.982Z","headline":"Kimi 团队开源 AgentENV：为 2.8 万亿参数模型提供智能体强化学习训练","description":"Moonshot AI 的 Kimi 团队与 kvcache-ai 开源了 AgentENV （AENV），一个用于大规模运行智能体环境的分布式平台，支持 Kimi K3 模型的智能体强化学习训练。该平台基于 Firecracker 微虚拟机，可实现快照启动/恢复低于 50 毫秒，并支持将运行中的沙箱分叉出最多 16 个独立子沙箱。代码采用 MIT 许可证。","url":"https://www.aioga.com/news/cms3pkqvt0105ro82u9dgd8hz/","mainEntityOfPage":"https://www.aioga.com/news/cms3pkqvt0105ro82u9dgd8hz/","datePublished":"2026-07-27T20:48:53.000Z","dateModified":"2026-07-27T20:48:53.000Z","inLanguage":"zh-CN","publisher":{"@type":"NewsMediaOrganization","name":"Aioga","url":"https://www.aioga.com"},"citation":["https://www.marktechpost.com/2026/07/27/kimi-ai-and-kvcache-ai-open-sources-agentenv","https://aihot.virxact.com/items/cms3pkqvt0105ro82u9dgd8hz"],"canonicalUrl":"https://www.aioga.com/news/cms3pkqvt0105ro82u9dgd8hz/","directAnswer":{"@type":"Answer","text":"Aioga 编辑摘要：Moonshot AI 的 Kimi 团队与 kvcache-ai 开源了 AgentENV （AENV），一个用于大规模运行智能体环境的分布式平台，支持 Kimi K3 模型的智能体强化学习训练。 Aioga 将其归入「产品更新」方向，重点关注它对真实使用和行业竞争的影响。","url":"https://www.aioga.com/news/cms3pkqvt0105ro82u9dgd8hz/","dateCreated":"2026-07-27T20:48:53.000Z","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":"marktechpost.com source article","url":"https://www.marktechpost.com/2026/07/27/kimi-ai-and-kvcache-ai-open-sources-agentenv","datePublished":"2026-07-27T20:48:53.000Z","provider":{"@type":"Organization","name":"marktechpost.com","url":"https://www.marktechpost.com/2026/07/27/kimi-ai-and-kvcache-ai-open-sources-agentenv"}},{"@type":"CreativeWork","name":"AIHot archive record","url":"https://aihot.virxact.com/items/cms3pkqvt0105ro82u9dgd8hz","datePublished":"2026-07-27T20:48:53.000Z","provider":{"@type":"Organization","name":"AIHot","url":"https://aihot.virxact.com/items/cms3pkqvt0105ro82u9dgd8hz"}}],"aggregationSource":"MarkTechPost（RSS）","originalPublisher":{"name":"marktechpost.com","url":"https://www.marktechpost.com/2026/07/27/kimi-ai-and-kvcache-ai-open-sources-agentenv"},"article":{"id":"cms3pkqvt0105ro82u9dgd8hz","slug":"cms3pkqvt0105ro82u9dgd8hz","url":"https://www.aioga.com/news/cms3pkqvt0105ro82u9dgd8hz/","title":"Kimi 团队开源 AgentENV：为 2.8 万亿参数模型提供智能体强化学习训练","title_en":"Kimi AI and kvcache-ai Open Sources 'AgentENV'： A Distributed System that Powers Agentic Reinforcement Learning （RL） Training for Kimi K3","summary":"Moonshot AI 的 Kimi 团队与 kvcache-ai 开源了 AgentENV （AENV），一个用于大规模运行智能体环境的分布式平台，支持 Kimi K3 模型的智能体强化学习训练。该平台基于 Firecracker 微虚拟机，可实现快照启动/恢复低于 50 毫秒，并支持将运行中的沙箱分叉出最多 16 个独立子沙箱。代码采用 MIT 许可证。","source":"MarkTechPost（RSS）","sourceUrl":"https://www.marktechpost.com/2026/07/27/kimi-ai-and-kvcache-ai-open-sources-agentenv","aiHotUrl":"https://aihot.virxact.com/items/cms3pkqvt0105ro82u9dgd8hz","publishedAt":"2026-07-27T20:48:53.000Z","category":"产品更新","score":54,"selected":false,"articleBody":["Moonshot AI’s Kimi team and kvcache-ai：https://github.com/kvcache-ai have open-sourced AgentENV (AENV)：https://github.com/kvcache-ai/AgentENV , a distributed platform for running agent environments at scale. AgentENV powers agentic reinforcement learning (RL) training for Kimi K3 , Moonshot’s 2.8-trillion-parameter Mixture-of-Experts model. The code ships under an MIT license .","Agentic RL does not just sample text. It requires the model to act inside a real computer. Every rollout needs an isolated Linux environment with a filesystem, a network stack, and live processes.","That requirement creates a hard trade-off. Containers start fast but share the host kernel, which weakens isolation for model-generated code. Full virtual machines isolate properly but boot slowly and hold memory while idle.","AgentENV targets exactly that gap. It runs Firecracker：https://github.com/firecracker-microvm/firecracker microVMs, then makes idle, restart, and branching cheap enough to run at training scale.","Each sandbox is a Firecracker microVM with its own Linux kernel, filesystem, and network namespace. Requests hit an Axum HTTP API, which forwards to an orchestrator that manages the sandbox lifecycle.","Storage is where the design gets interesting. The rootfs is served through a ublk userspace block device backed by overlaybd：https://github.com/containerd/overlaybd layered images. Read-only base layers are shared across sandboxes, and each sandbox writes into its own upper layer.","Inside every guest, a daemon called envd handles command execution, file operations, and health reporting on port 49983 . A reverse proxy routes HTTP and WebSocket traffic from clients to services running inside the VM.","The project also lists two density mechanisms. The host page cache is shared across storage and memory-snapshot data. Memory ballooning returns reclaimable guest memory to the host, which sustains overcommit as environments diverge over time.","These four features are the reason the project exists. AgentENV snapshots memory and filesystem changes incrementally rather than writing a full image each time.","The reported figures are as follows. Snapshot-backed environments boot or resume in under 50 ms and pause in under 100 ms . Incremental snapshot capture completes in under 100 ms , even under heavy disk modification.","Fork is the feature most specific to RL. A running sandbox can clone into up to 16 independent child sandboxes on the same node. The source pauses briefly during capture, then resumes. Every child inherits the source filesystem, memory, and resource configuration.","The practical effect is that expensive setup runs once. A team can install dependencies, clone a repository, and reach a task state. That exact state then branches into parallel rollouts. Snapshots persist to S3-compatible object storage or a shared distributed filesystem.","One default is worth noting. Every sandbox carries a TTL, and expiry triggers a pause , not a delete. Deletion requires passing autoPause: false on the create API.","Images load on demand through overlaybd. Local disk acts as a bounded cache that retains hot data and evicts cold data.","This is what makes the fleet-level story work. Nodes do not need to pre-warm every image or hold a complete copy of every snapshot. The addressable image set can therefore exceed local disk capacity while startup stays fast cluster-wide.","Snapshot state is organized in three layers. A builder staging workspace holds artifacts during a build. A committed snapshot repository is the durable source of truth. A node-local runtime cache holds launch-time derived configs.","Two repository backends are supported: posix_fs (default) and oss . The oss path runs through a shared S3-compatible client, so an explicit region is required.","An optional peer-to-peer transport based on iroh can advertise committed artifacts to peer nodes. It is disabled by default . The documentation is explicit that P2P does not change the committed snapshot model. For shared storage, the docs ask for at least 1 Gbps and strongly recommend 10 Gbps or faster.","AgentENV exposes an E2B-compatible：https://kvcache-ai.github.io/AgentENV/integration/e2b.html HTTP API . Point E2B_API_URL at your server and the official E2B：https://github.com/e2b-dev/e2b Python or TypeScript SDK works without code changes.","This is a deliberate distribution choice. Teams already running agents on E2B can self-host the runtime without rewriting their agent code. A native aenv CLI is also provided, and the docs recommend it for AgentENV-specific workflows.","Prerequisites are Linux kernel 6.8+ and /dev/kvm access; the install script additionally requires Ubuntu 24.04. The aenv CLI supports Linux and macOS on x86_64 and arm64. The server itself is Linux-only, because it needs KVM.","Multi-node deployments add a gateway on :8080 and a scheduler on :9090 .","Check out the GitHub repository：https://github.com/kvcache-ai/AgentENV and documentation：https://kvcache-ai.github.io/AgentENV/. All credit for this research goes to the researchers of this project.","Asif Razzaq is the CEO of Marktechpost Media Inc.. As a visionary entrepreneur and engineer, Asif is committed to harnessing the potential of Artificial Intelligence for social good. His most recent endeavor is the launch of an Artificial Intelligence Media Platform, Marktechpost, which stands out for its in-depth coverage of machine learning and deep learning news that is both technically sound and easily understandable by a wide audience. The platform boasts of over 2 million monthly views, illustrating its popularity among audiences.","Build an Agentic Event Venue Operator [Full Codes]：https://pxllnk.co/twdn5","Thanks! Our team will contact you soon 🙌"],"articleImages":[{"sourceUrl":"https://www.marktechpost.com/wp-content/uploads/2019/06/Screen-Shot-2021-09-14-at-9.02.24-AM-300x300.png","alt":"","afterParagraph":22,"url":"/media/articles/cms3pkqvt0105ro82u9dgd8hz/787a6d54564e8e19.webp"},{"sourceUrl":"https://www.marktechpost.com/wp-content/uploads/2026/07/blog6171-9-100x70.png","alt":"Designing Skill-Driven Financial Analysis Agents with Claude, Python, MCP Connectors, and Automated Deliverables","afterParagraph":23,"url":"/media/articles/cms3pkqvt0105ro82u9dgd8hz/3c7ad2a86fecc3da.png"},{"sourceUrl":"https://www.marktechpost.com/wp-content/uploads/2026/07/blog6171-8-100x70.png","alt":"How Perplexity's pplx CLI Gives Coding Agents Live Web Search From the Command Line","afterParagraph":23,"url":"/media/articles/cms3pkqvt0105ro82u9dgd8hz/8cdd9bfb6f119048.png"},{"sourceUrl":"https://www.marktechpost.com/wp-content/uploads/2026/07/blog6171-6-100x70.png","alt":"KwaiKAT Team Releases KAT-Coder-V2.5","afterParagraph":23,"url":"/media/articles/cms3pkqvt0105ro82u9dgd8hz/2225a1a0a768ea5a.webp"}],"mediaStatus":"ok","articleBodyZh":["Moonshot AI 的 Kimi 团队和 kvcache-ai：https://github.com/kvcache-ai 已开源了 AgentENV (AENV)：https://github.com/kvcache-ai/AgentENV，这是一个用于大规模运行代理环境的分布式平台。AgentENV 支持 Kimi K3 的代理强化学习（RL）训练，Kimi K3 是 Moonshot 的 2.8 万亿参数专家混合模型（Mixture-of-Experts）。该代码采用 MIT 许可证发布。","代理强化学习不仅仅是生成文本。它要求模型在真实计算机内执行操作。每一次回滚都需要一个带有文件系统、网络堆栈和活动进程的隔离 Linux 环境。","这一要求带来了一个难以权衡的选择。容器启动快，但共享主机内核，这削弱了对模型生成代码的隔离。完整虚拟机可以正确隔离，但启动慢，空闲时仍占用内存。","AgentENV 正是针对这一空缺设计的。它运行 Firecracker：https://github.com/firecracker-microvm/firecracker 微型虚拟机，然后使空闲、重启和分支操作足够便宜，从而可以在训练规模下运行。","每个沙箱都是一个带有独立 Linux 内核、文件系统和网络命名空间的 Firecracker 微型虚拟机。请求通过 Axum HTTP API 接入，然后转发给管理沙箱生命周期的调度器。","存储设计非常有趣。根文件系统通过由 overlaybd：https://github.com/containerd/overlaybd 分层镜像支持的 ublk 用户空间块设备提供。只读基础层在沙箱间共享，每个沙箱将数据写入其自己的上层。","在每个客户机内，一个名为 envd 的守护进程负责命令执行、文件操作和在端口 49983 上的健康报告。反向代理将客户端的 HTTP 和 WebSocket 流量路由到 VM 内运行的服务。","该项目还列出了两种密度机制。主机页缓存跨存储和内存快照数据共享。内存气球技术将可回收的客户机内存返回给主机，从而在环境随时间变化时维持过量分配。","这四个功能是该项目存在的原因。AgentENV 以增量方式快照内存和文件系统的变化，而不是每次都写入完整镜像。","报告的数字如下。以快照为基础的环境启动或恢复时间不到50毫秒，暂停时间不到100毫秒。在磁盘修改繁重的情况下，增量快照捕获也能在100毫秒内完成。","Fork 是 RL 最独特的功能。运行中的沙箱可以在同一节点上克隆成最多16个独立的子沙箱。源沙箱在捕获期间会短暂暂停，然后恢复。每个子沙箱继承源沙箱的文件系统、内存和资源配置。","实际效果是昂贵的设置只需运行一次。团队可以安装依赖项、克隆代码库并达到任务状态。然后，该确切状态可以分支为并行部署。快照会持久化到兼容 S3 的对象存储或共享分布式文件系统中。","有一个默认值值得注意。每个沙箱都有 TTL，过期会触发暂停，而非删除。删除需要在创建 API 中传递 autoPause: false。","镜像通过 overlaybd 按需加载。本地磁盘充当有界缓存，保留热数据并驱逐冷数据。","这就是使集群级方案可行的原因。节点不需要预热每个镜像或保留每个快照的完整副本。因此，可寻址的镜像集合可以超过本地磁盘容量，同时启动在整个集群中保持快速。","快照状态分为三层。构建器暂存工作区在构建过程中保存工件。已提交的快照仓库是持久化的真实来源。节点本地运行时缓存保存启动时派生的配置。","支持两种仓库后端：posix_fs（默认）和 oss。oss 路径通过共享的 S3 兼容客户端运行，因此需要指定区域。","可选的基于 iroh 的点对点传输可以向同伴节点公布已提交的工件。默认情况下禁用。文档明确指出 P2P 不会改变已提交的快照模型。对于共享存储，文档要求至少 1 Gbps，并强烈推荐 10 Gbps 或更快。","AgentENV 提供 E2B 兼容的：https://kvcache-ai.github.io/AgentENV/integration/e2b.html HTTP API。将 E2B_API_URL 指向你的服务器，官方 E2B：https://github.com/e2b-dev/e2b Python 或 TypeScript SDK 可在不更改代码的情况下使用。","这是一个有意的分发选择。已经在 E2B 上运行代理的团队可以自托管运行时，而无需重写他们的代理代码。还提供了一个原生的 aenv CLI，文档推荐在 AgentENV 特定工作流程中使用它。","先决条件是 Linux 内核 6.8 及以上以及 /dev/kvm 访问；安装脚本还需要 Ubuntu 24.04。aenv CLI 支持 x86_64 和 arm64 架构的 Linux 和 macOS。服务器本身仅支持 Linux，因为它需要 KVM。","多节点部署在 :8080 添加一个网关，在 :9090 添加一个调度器。","请查看 GitHub 仓库：https://github.com/kvcache-ai/AgentENV 以及文档：https://kvcache-ai.github.io/AgentENV/。所有研究成果归此项目研究人员所有。","Asif Razzaq 是 Marktechpost Media Inc. 的首席执行官。作为一名富有远见的企业家和工程师，Asif 致力于利用人工智能的潜力造福社会。他最近的努力是推出人工智能媒体平台 Marktechpost，该平台因其对机器学习和深度学习新闻的深入报道而脱颖而出，这些报道既技术可靠，又易于广大读者理解。该平台月访问量超过 200 万，显示出其在观众中的受欢迎程度。","构建一个代理事件场馆运营商 [完整代码]：https://pxllnk.co/twdn5","谢谢！我们的团队将很快与您联系 🙌"],"translationStatus":"translated","bodyOrigin":"source-page","editorial":{"summary":"Aioga 编辑摘要：Moonshot AI 的 Kimi 团队与 kvcache-ai 开源了 AgentENV （AENV），一个用于大规模运行智能体环境的分布式平台，支持 Kimi K3 模型的智能体强化学习训练。 Aioga 将其归入「产品更新」方向，重点关注它对真实使用和行业竞争的影响。","background":"背景分析：模型与研究类动态需要结合能力边界、开放方式、成本、可用性和真实任务表现判断，单项指标领先不等于已经形成稳定采用。","viewpoint":"Aioga 判断：这条动态更适合作为行业观察信号，当前信息足以建立线索，但不足以推导长期结论。","implications":"影响分析：对相关团队而言，短期应先核对来源、可用范围和实际成本，再判断是否值得接入或跟进。","nextStep":"后续观察：继续观察官方文档、实际可用性、价格变化、开发者反馈和竞品回应。","evidenceRefs":["title","summary","articleBody"],"confidence":"medium","status":"published","aiGenerated":false,"autoApproved":true,"generatedBy":"rule-safe-fallback","generatedAt":"2026-07-28T05:28:54.610Z","sourceHash":"26c49f17ea5dc5bf","validation":{"passed":true,"mode":"rule-safe-fallback","checks":["schema","length","source-attribution","no-html"]}},"tags":["产品更新","MarkTechPost（RSS）"],"translations":{"zh-CN":{"title":"Kimi 团队开源 AgentENV：为 2.8 万亿参数模型提供智能体强化学习训练","summary":"Moonshot AI 的 Kimi 团队与 kvcache-ai 开源了 AgentENV （AENV），一个用于大规模运行智能体环境的分布式平台，支持 Kimi K3 模型的智能体强化学习训练。该平台基于 Firecracker 微虚拟机，可实现快照启动/恢复低于 50 毫秒，并支持将运行中的沙箱分叉出最多 16 个独立子沙箱。代码采用 MIT 许可证。","category":"产品更新","source":"marktechpost.com","aggregationSource":"MarkTechPost（RSS）","pageTitle":"Kimi 团队开源 AgentENV：为 2.8 万亿参数模型提供智能体强化学习训练 - Aioga AI资讯","description":"Moonshot AI 的 Kimi 团队与 kvcache-ai 开源了 AgentENV （AENV），一个用于大规模运行智能体环境的分布式平台，支持 Kimi K3 模型的智能体强化学习训练。该平台基于 Firecracker 微虚拟机，可实现快照启动/恢复低于 50 毫秒，并支持将运行中的沙箱分叉出最多 16 个独立子沙箱。代码采用 MIT 许可证。","url":"https://www.aioga.com/news/cms3pkqvt0105ro82u9dgd8hz/"},"en":{"title":"Kimi team open-sources AgentENV: providing agent reinforcement learning training for 2.8 trillion-parameter models","summary":"The Kimi team at Moonshot AI and kvcache-ai have open-sourced AgentENV (AENV), a distributed platform for running agent environments at scale, supporting reinforcement learning training for agents using the Kimi K3 model. The platform is based on Firecracker microVMs, enabling snapshot start/resume in under 50 milliseconds, and supports forking running sandboxes into up to 16 independent sub-sandboxes. The code is licensed under the MIT license.","category":"Products","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"Kimi team open-sources AgentENV: providing agent reinforcement learning training for 2.8 trillion-parameter models - Aioga AI News","description":"The Kimi team at Moonshot AI and kvcache-ai have open-sourced AgentENV (AENV), a distributed platform for running agent environments at scale, supporting reinforcement learning tra...","url":"https://www.aioga.com/en/news/cms3pkqvt0105ro82u9dgd8hz/","contentTranslated":true,"sourceHash":"629af90d2d1ea50d","translatedAt":"2026-07-27T21:43:04.556Z"},"ja":{"title":"Kimi チームのオープンソース AgentENV：2.8 兆パラメータモデル向けのエージェント強化学習トレーニングを提供","summary":"Moonshot AI の Kimi チームは kvcache-ai と共に AgentENV（AENV）をオープンソース化しました。これは大規模にエージェント環境を運用するための分散プラットフォームで、Kimi K3 モデルのエージェント強化学習トレーニングをサポートします。このプラットフォームは Firecracker マイクロ仮想マシンに基づいており、スナップショットの起動/復元を 50 ミリ秒未満で実現でき、実行中のサンドボックスを最大 16 個の独立したサブサンドボックスに分岐することをサポートします。コードは MIT ライセンスを採用しています。","category":"製品更新","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"Kimi チームのオープンソース AgentENV：2.8 兆パラメータモデル向けのエージェント強化学習トレーニングを提供 - Aioga AIニュース","description":"Moonshot AI の Kimi チームは kvcache-ai と共に AgentENV（AENV）をオープンソース化しました。これは大規模にエージェント環境を運用するための分散プラットフォームで、Kimi K3 モデルのエージェント強化学習トレーニングをサポートします。このプラットフォームは Firecracker マイクロ仮想マシンに基づいており、...","url":"https://www.aioga.com/ja/news/cms3pkqvt0105ro82u9dgd8hz/","contentTranslated":true,"sourceHash":"629af90d2d1ea50d","translatedAt":"2026-07-27T21:43:17.776Z"},"ko":{"title":"Kimi 팀 오픈소스 AgentENV: 2.8조 매개변수 모델을 위한 에이전트 강화 학습 훈련 제공","summary":"Moonshot AI의 Kimi 팀이 kvcache-ai와 함께 AgentENV(AENV)를 오픈소스화했으며, 이는 대규모 에이전트 환경 실행을 위한 분산 플랫폼으로 Kimi K3 모델의 에이전트 강화 학습 훈련을 지원합니다. 이 플랫폼은 Firecracker 마이크로 가상 머신을 기반으로 하며, 스냅샷 시작/복구를 50밀리초 이하로 구현할 수 있고, 실행 중인 샌드박스를 최대 16개의 독립 하위 샌드박스로 분기하는 것을 지원합니다. 코드는 MIT 라이선스를 채택하고 있습니다.","category":"제품 업데이트","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"Kimi 팀 오픈소스 AgentENV: 2.8조 매개변수 모델을 위한 에이전트 강화 학습 훈련 제공 - Aioga AI 뉴스","description":"Moonshot AI의 Kimi 팀이 kvcache-ai와 함께 AgentENV(AENV)를 오픈소스화했으며, 이는 대규모 에이전트 환경 실행을 위한 분산 플랫폼으로 Kimi K3 모델의 에이전트 강화 학습 훈련을 지원합니다. 이 플랫폼은 Firecracker 마이크로 가상 머신을 기반으로 하며, 스냅샷 시작/복구를 50...","url":"https://www.aioga.com/ko/news/cms3pkqvt0105ro82u9dgd8hz/","contentTranslated":true,"sourceHash":"629af90d2d1ea50d","translatedAt":"2026-07-27T21:43:59.598Z"},"es":{"title":"Kimi equipo lanza AgentENV de código abierto: proporciona entrenamiento de aprendizaje por refuerzo de agentes para modelos de 2.8 billones de parámetros","summary":"El equipo Kimi de Moonshot AI y kvcache-ai lanzaron AgentENV (AENV), una plataforma distribuida para ejecutar entornos de agentes a gran escala, que admite el entrenamiento de aprendizaje por refuerzo de agentes con el modelo Kimi K3. La plataforma se basa en la microvirtualización Firecracker, lo que permite iniciar/restaurar instantáneas en menos de 50 milisegundos y admite bifurcar hasta 16 sub-sandbox independientes desde un sandbox en ejecución. El código está bajo la licencia MIT.","category":"Productos","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"Kimi equipo lanza AgentENV de código abierto: proporciona entrenamiento de aprendizaje por refuerzo de agentes para modelos de 2.8 billones de parámetros - Aioga Noticias de IA","description":"El equipo Kimi de Moonshot AI y kvcache-ai lanzaron AgentENV (AENV), una plataforma distribuida para ejecutar entornos de agentes a gran escala, que admite el entrenamiento de apre...","url":"https://www.aioga.com/es/news/cms3pkqvt0105ro82u9dgd8hz/","contentTranslated":true,"sourceHash":"629af90d2d1ea50d","translatedAt":"2026-07-27T21:43:57.040Z"},"fr":{"title":"L'équipe Kimi publie en open source AgentENV : entraînement d'agents par apprentissage par renforcement pour des modèles à 2,8 billions de paramètres","summary":"L'équipe Kimi de Moonshot AI et kvcache-ai ont open-sourcé AgentENV (AENV), une plateforme distribuée pour exécuter des environnements d'agents à grande échelle, supportant l'entraînement d'agents en apprentissage par renforcement avec le modèle Kimi K3. La plateforme est basée sur des micro-machines virtuelles Firecracker, permettant un démarrage/reprise rapide en moins de 50 millisecondes, et supporte la duplication d'un bac à sable en cours d'exécution jusqu'à 16 sous-bacs à sable indépendants. Le code est sous licence MIT.","category":"Produits","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"L'équipe Kimi publie en open source AgentENV : entraînement d'agents par apprentissage par renforcement pour des modèles à 2,8 billions de paramètres - Aioga Actualités IA","description":"L'équipe Kimi de Moonshot AI et kvcache-ai ont open-sourcé AgentENV (AENV), une plateforme distribuée pour exécuter des environnements d'agents à grande échelle, supportant l'entra...","url":"https://www.aioga.com/fr/news/cms3pkqvt0105ro82u9dgd8hz/","contentTranslated":true,"sourceHash":"629af90d2d1ea50d","translatedAt":"2026-07-27T21:44:35.784Z"},"de":{"title":"Kimi-Team veröffentlicht Open-Source AgentENV: Bietet Agenten-RL-Training für Modelle mit 2,8 Billionen Parametern","summary":"Das Kimi-Team von Moonshot AI hat zusammen mit kvcache-ai AgentENV (AENV) als Open-Source veröffentlicht, eine verteilte Plattform für den großflächigen Betrieb von Agenten-Umgebungen, die das Reinforcement Learning Training von Agenten mit dem Kimi K3 Modell unterstützt. Die Plattform basiert auf der Firecracker-Mikrovirtualmaschine, ermöglicht Snapshots zum Starten/Wiederherstellen in unter 50 Millisekunden und unterstützt das Abzweigen von bis zu 16 separaten Unter-Sandboxen von laufenden Sandboxes. Der Code steht unter der MIT-Lizenz.","category":"产品更新","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"Kimi-Team veröffentlicht Open-Source AgentENV: Bietet Agenten-RL-Training für Modelle mit 2,8 Billionen Parametern - Aioga KI-News","description":"Das Kimi-Team von Moonshot AI hat zusammen mit kvcache-ai AgentENV (AENV) als Open-Source veröffentlicht, eine verteilte Plattform für den großflächigen Betrieb von Agenten-Umgebun...","url":"https://www.aioga.com/de/news/cms3pkqvt0105ro82u9dgd8hz/","contentTranslated":true,"sourceHash":"629af90d2d1ea50d","translatedAt":"2026-07-27T21:44:34.328Z"},"pt-BR":{"title":"Equipe Kimi lança AgentENV de código aberto: treinamento de aprendizado por reforço de agente para modelos de 2,8 trilhões de parâmetros","summary":"A equipe Kimi da Moonshot AI e a kvcache-ai lançaram o AgentENV (AENV), uma plataforma distribuída para executar ambientes de agentes em grande escala, suportando o treinamento de aprendizado por reforço de agentes com o modelo Kimi K3. A plataforma é baseada em micro máquinas virtuais Firecracker, podendo realizar inicialização/recuperação de snapshots em menos de 50 milissegundos, e suporta o desdobramento de até 16 sub-sandbox independentes a partir de um sandbox em execução. O código utiliza a licença MIT.","category":"产品更新","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"Equipe Kimi lança AgentENV de código aberto: treinamento de aprendizado por reforço de agente para modelos de 2,8 trilhões de parâmetros - Aioga Notícias de IA","description":"A equipe Kimi da Moonshot AI e a kvcache-ai lançaram o AgentENV (AENV), uma plataforma distribuída para executar ambientes de agentes em grande escala, suportando o treinamento de...","url":"https://www.aioga.com/pt-BR/news/cms3pkqvt0105ro82u9dgd8hz/","contentTranslated":true,"sourceHash":"629af90d2d1ea50d","translatedAt":"2026-07-27T21:45:20.344Z"},"ru":{"title":"Команда Kimi открыла исходный код AgentENV: обучение агентов с подкреплением для модели с 2,8 триллиона параметров","summary":"Команда Kimi компании Moonshot AI вместе с kvcache-ai выпустила AgentENV (AENV), распределённую платформу для масштабного запуска сред для агентов, поддерживающую обучение с подкреплением агентов модели Kimi K3. Платформа основана на микро-ВМ Firecracker, что позволяет запускать/восстанавливать снимки за менее чем 50 миллисекунд, и поддерживает форк существующего песочницы до 16 отдельных под-песочниц. Код распространяется под лицензией MIT.","category":"产品更新","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"Команда Kimi открыла исходный код AgentENV: обучение агентов с подкреплением для модели с 2,8 триллиона параметров - Aioga Новости ИИ","description":"Команда Kimi компании Moonshot AI вместе с kvcache-ai выпустила AgentENV (AENV), распределённую платформу для масштабного запуска сред для агентов, поддерживающую обучение с подкре...","url":"https://www.aioga.com/ru/news/cms3pkqvt0105ro82u9dgd8hz/","contentTranslated":true,"sourceHash":"629af90d2d1ea50d","translatedAt":"2026-07-27T21:45:12.351Z"},"ar":{"title":"فريق Kimi يفتح مصدر AgentENV: لتوفير تدريب التعلم المعزز للوكيل لنموذج يحتوي على 2.8 تريليون معلمة","summary":"أطلقت فرق Kimi التابعة لشركة Moonshot AI و kvcache-ai منصة AgentENV (AENV) مفتوحة المصدر، وهي منصة موزعة لتشغيل بيئات الوكلاء على نطاق واسع، وتدعم تدريب التعلم المعزز لوكلاء نموذج Kimi K3. تعتمد هذه المنصة على آلة Firecracker الدقيقة الافتراضية، ويمكنها تنفيذ بدء التشغيل/الاستعادة من اللقطات في أقل من 50 مللي ثانية، وتدعم إنشاء ما يصل إلى 16 صندوقًا فرعيًا مستقلاً من الصندوق الجاري تشغيله. الشيفرة مرخصة بموجب رخصة MIT.","category":"产品更新","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"فريق Kimi يفتح مصدر AgentENV: لتوفير تدريب التعلم المعزز للوكيل لنموذج يحتوي على 2.8 تريليون معلمة - Aioga أخبار الذكاء الاصطناعي","description":"أطلقت فرق Kimi التابعة لشركة Moonshot AI و kvcache-ai منصة AgentENV (AENV) مفتوحة المصدر، وهي منصة موزعة لتشغيل بيئات الوكلاء على نطاق واسع، وتدعم تدريب التعلم المعزز لوكلاء نموذج...","url":"https://www.aioga.com/ar/news/cms3pkqvt0105ro82u9dgd8hz/","contentTranslated":true,"sourceHash":"629af90d2d1ea50d","translatedAt":"2026-07-27T21:45:59.383Z"},"hi":{"title":"Kimi टीम ने AgentENV ओपन सोर्स किया: 2.8 ट्रिलियन पैरामीटर मॉडल के लिए एजेंट सशक्तिकरण सीखने का प्रशिक्षण प्रदान करता है","summary":"Moonshot AI की Kimi टीम ने kvcache-ai के साथ मिलकर AgentENV (AENV) ओपन-सोर्स किया, जो बड़े पैमाने पर एजेंट वातावरण चलाने के लिए एक वितरित प्लेटफ़ॉर्म है, और Kimi K3 मॉडल के एजेंट रिइन्फोर्समेंट लर्निंग प्रशिक्षण का समर्थन करता है। यह प्लेटफ़ॉर्म Firecracker माइक्रो वर्चुअल मशीन पर आधारित है, जो स्नैपशॉट स्टार्ट/रिकवरी को 50 मिलीसेकंड से कम समय में सक्षम बनाता है, और चल रहे सैंडबॉक्स से अधिकतम 16 स्वतंत्र उप-सैंडबॉक्स फोर्क करने का समर्थन करता है। कोड MIT लाइसेंस के तहत है।","category":"产品更新","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"Kimi टीम ने AgentENV ओपन सोर्स किया: 2.8 ट्रिलियन पैरामीटर मॉडल के लिए एजेंट सशक्तिकरण सीखने का प्रशिक्षण प्रदान करता है - Aioga AI समाचार","description":"Moonshot AI की Kimi टीम ने kvcache-ai के साथ मिलकर AgentENV (AENV) ओपन-सोर्स किया, जो बड़े पैमाने पर एजेंट वातावरण चलाने के लिए एक वितरित प्लेटफ़ॉर्म है, और Kimi K3 मॉडल के एजेंट र...","url":"https://www.aioga.com/hi/news/cms3pkqvt0105ro82u9dgd8hz/","contentTranslated":true,"sourceHash":"629af90d2d1ea50d","translatedAt":"2026-07-27T21:45:58.750Z"},"it":{"title":"Kimi team rilascia open source AgentENV: addestramento di agenti con apprendimento rinforzato per modelli da 2,8 trilioni di parametri","summary":"Il team Kimi di Moonshot AI e kvcache-ai hanno reso open source AgentENV (AENV), una piattaforma distribuita per l'esecuzione su larga scala di ambienti per agenti, che supporta l'addestramento di agenti con apprendimento rinforzato del modello Kimi K3. La piattaforma è basata sulle micro-virtual machine Firecracker, consentendo un avvio/ripristino con snapshot inferiore a 50 millisecondi, e supporta la divisione di sandbox in esecuzione fino a un massimo di 16 sotto-sandbox indipendenti. Il codice è rilasciato sotto licenza MIT.","category":"产品更新","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"Kimi team rilascia open source AgentENV: addestramento di agenti con apprendimento rinforzato per modelli da 2,8 trilioni di parametri - Aioga Notizie IA","description":"Il team Kimi di Moonshot AI e kvcache-ai hanno reso open source AgentENV (AENV), una piattaforma distribuita per l'esecuzione su larga scala di ambienti per agenti, che supporta l'...","url":"https://www.aioga.com/it/news/cms3pkqvt0105ro82u9dgd8hz/","contentTranslated":true,"sourceHash":"629af90d2d1ea50d","translatedAt":"2026-07-27T21:46:45.394Z"},"nl":{"title":"Kimi-team open source AgentENV: biedt agentversterkend leren training voor modellen met 2,8 biljoen parameters","summary":"Het Kimi-team van Moonshot AI en kvcache-ai hebben AgentENV (AENV) open source gemaakt, een gedistribueerd platform voor het grootschalig uitvoeren van agent-omgevingen, dat de training van agenten met versterkend leren ondersteunt voor het Kimi K3-model. Het platform is gebaseerd op Firecracker-microvirtuele machines en kan snapshots starten/herstellen in minder dan 50 milliseconden, en ondersteunt het splitsen van een draaiende sandbox in maximaal 16 afzonderlijke sub-sandboxes. De code is gelicentieerd onder de MIT-licentie.","category":"产品更新","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"Kimi-team open source AgentENV: biedt agentversterkend leren training voor modellen met 2,8 biljoen parameters - Aioga AI-nieuws","description":"Het Kimi-team van Moonshot AI en kvcache-ai hebben AgentENV (AENV) open source gemaakt, een gedistribueerd platform voor het grootschalig uitvoeren van agent-omgevingen, dat de tra...","url":"https://www.aioga.com/nl/news/cms3pkqvt0105ro82u9dgd8hz/","contentTranslated":true,"sourceHash":"629af90d2d1ea50d","translatedAt":"2026-07-27T21:46:40.022Z"},"tr":{"title":"Kimi takım açık kaynak AgentENV: 2,8 trilyon parametreli model için yapay zekâ güçlendirme öğrenimi eğitimi sağlıyor","summary":"Moonshot AI'nin Kimi ekibi, kvcache-ai ile birlikte AgentENV (AENV) adlı, büyük ölçekli ajan ortamları çalıştırmak için kullanılan bir dağıtılmış platformu açık kaynak yaptı ve Kimi K3 modeli ajanlarının takviye öğrenme eğitimini destekliyor. Bu platform, Firecracker mikro sanal makinelerine dayanıyor, 50 milisaniyeden daha kısa sürede anlık görüntü başlatma/gözden geçirme sağlayabiliyor ve çalışmakta olan sandbox'ın en fazla 16 bağımsız alt sandbox'a çatallanmasını destekliyor. Kod MIT lisansı ile lisanslanmıştır.","category":"产品更新","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"Kimi takım açık kaynak AgentENV: 2,8 trilyon parametreli model için yapay zekâ güçlendirme öğrenimi eğitimi sağlıyor - Aioga AI Haberleri","description":"Moonshot AI'nin Kimi ekibi, kvcache-ai ile birlikte AgentENV (AENV) adlı, büyük ölçekli ajan ortamları çalıştırmak için kullanılan bir dağıtılmış platformu açık kaynak yaptı ve Kim...","url":"https://www.aioga.com/tr/news/cms3pkqvt0105ro82u9dgd8hz/","contentTranslated":true,"sourceHash":"629af90d2d1ea50d","translatedAt":"2026-07-27T21:47:23.839Z"},"vi":{"title":"Nhóm Kimi phát hành AgentENV mã nguồn mở: Cung cấp huấn luyện học tăng cường agent cho mô hình 2,8 nghìn tỷ tham số","summary":"Nhóm Kimi của Moonshot AI cùng với kvcache-ai đã phát hành mã nguồn mở AgentENV (AENV), một nền tảng phân tán để chạy môi trường agent ở quy mô lớn, hỗ trợ đào tạo học tăng cường agent với mô hình Kimi K3. Nền tảng này dựa trên máy ảo vi mô Firecracker, có thể khởi động/lấy lại snapshot dưới 50 mili giây, và hỗ trợ phân nhánh hộp cát đang chạy thành tối đa 16 hộp cát con độc lập. Mã nguồn được cấp phép theo MIT.","category":"产品更新","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"Nhóm Kimi phát hành AgentENV mã nguồn mở: Cung cấp huấn luyện học tăng cường agent cho mô hình 2,8 nghìn tỷ tham số - Tin tức AI Aioga","description":"Nhóm Kimi của Moonshot AI cùng với kvcache-ai đã phát hành mã nguồn mở AgentENV (AENV), một nền tảng phân tán để chạy môi trường agent ở quy mô lớn, hỗ trợ đào tạo học tăng cường a...","url":"https://www.aioga.com/vi/news/cms3pkqvt0105ro82u9dgd8hz/","contentTranslated":true,"sourceHash":"629af90d2d1ea50d","translatedAt":"2026-07-27T21:47:23.111Z"},"id":{"title":"Tim Kimi merilis AgentENV sumber terbuka: menyediakan pelatihan pembelajaran penguatan agen untuk model 2,8 triliun parameter","summary":"Tim Kimi dari Moonshot AI bersama kvcache-ai telah merilis AgentENV (AENV), sebuah platform terdistribusi untuk menjalankan lingkungan agen berskala besar, mendukung pelatihan pembelajaran penguatan agen dengan model Kimi K3. Platform ini berbasis mikrovirtual mesin Firecracker, mampu melakukan startup/pemulihan snapshot kurang dari 50 milidetik, dan mendukung pemisahan sandbox yang sedang berjalan menjadi hingga 16 sub-sandbox independen. Kode ini menggunakan lisensi MIT.","category":"产品更新","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"Tim Kimi merilis AgentENV sumber terbuka: menyediakan pelatihan pembelajaran penguatan agen untuk model 2,8 triliun parameter - Berita AI Aioga","description":"Tim Kimi dari Moonshot AI bersama kvcache-ai telah merilis AgentENV (AENV), sebuah platform terdistribusi untuk menjalankan lingkungan agen berskala besar, mendukung pelatihan pemb...","url":"https://www.aioga.com/id/news/cms3pkqvt0105ro82u9dgd8hz/","contentTranslated":true,"sourceHash":"629af90d2d1ea50d","translatedAt":"2026-07-27T21:48:02.121Z"},"th":{"title":"ทีม Kimi เปิดตัวซอฟต์แวร์ AgentENV แบบโอเพนซอร์ส: สำหรับการฝึกอบรมการเรียนรู้ด้วยการเสริมกำลังของเอเย่นต์สำหรับโมเดล 2.8 ล้านล้านพารามิเตอร์","summary":"ทีม Kimi ของ Moonshot AI และ kvcache-ai เปิดตัว AgentENV (AENV) ซึ่งเป็นแพลตฟอร์มแบบกระจายที่ใช้สำหรับการรันสภาพแวดล้อมของเอเย่นต์ขนาดใหญ่ สนับสนุนการฝึกการเรียนรู้แบบเสริมแรงของเอเย่นต์โดยใช้โมเดล Kimi K3 แพลตฟอร์มนี้ใช้ Firecracker ไมโครเวอร์ชวลแมชชีน สามารถเริ่ม/กู้คืนสแน็ปชอตได้ภายในเวลาไม่ถึง 50 มิลลิวินาที และสนับสนุนการแยกซ้ำของซองทรายที่กำลังรันออกเป็นซองทรายย่อยอิสระสูงสุด 16 ตัว โค้ดใช้ใบอนุญาต MIT","category":"产品更新","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"ทีม Kimi เปิดตัวซอฟต์แวร์ AgentENV แบบโอเพนซอร์ส: สำหรับการฝึกอบรมการเรียนรู้ด้วยการเสริมกำลังของเอเย่นต์สำหรับโมเดล 2.8 ล้านล้านพารามิเตอร์ - ข่าว AI Aioga","description":"ทีม Kimi ของ Moonshot AI และ kvcache-ai เปิดตัว AgentENV (AENV) ซึ่งเป็นแพลตฟอร์มแบบกระจายที่ใช้สำหรับการรันสภาพแวดล้อมของเอเย่นต์ขนาดใหญ่ สนับสนุนการฝึกการเรียนรู้แบบเสริมแรงของเอ...","url":"https://www.aioga.com/th/news/cms3pkqvt0105ro82u9dgd8hz/","contentTranslated":true,"sourceHash":"629af90d2d1ea50d","translatedAt":"2026-07-27T21:48:06.436Z"},"pl":{"title":"Zespół Kimi udostępnia AgentENV jako otwarte oprogramowanie: zapewnia trening wzmocnieniowy dla modeli o 2,8 biliona parametrów","summary":"Zespół Kimi z Moonshot AI wraz z kvcache-ai udostępnili AgentENV (AENV), rozproszoną platformę do masowego uruchamiania środowisk agentów, wspierającą trening wzmocnionej nauki agentów modelu Kimi K3. Platforma oparta jest na mikro-wirtualnej maszynie Firecracker, umożliwiając uruchamianie/przywracanie z migawki w czasie krótszym niż 50 milisekund, oraz obsługuje tworzenie maksymalnie 16 niezależnych podśrodowisk z działającego sandboxa. Kod objęty jest licencją MIT.","category":"产品更新","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"Zespół Kimi udostępnia AgentENV jako otwarte oprogramowanie: zapewnia trening wzmocnieniowy dla modeli o 2,8 biliona parametrów - Aioga Wiadomości AI","description":"Zespół Kimi z Moonshot AI wraz z kvcache-ai udostępnili AgentENV (AENV), rozproszoną platformę do masowego uruchamiania środowisk agentów, wspierającą trening wzmocnionej nauki age...","url":"https://www.aioga.com/pl/news/cms3pkqvt0105ro82u9dgd8hz/","contentTranslated":true,"sourceHash":"629af90d2d1ea50d","translatedAt":"2026-07-27T21:48:52.727Z"}}}}