{"id":24596,"date":"2025-07-12T05:21:34","date_gmt":"2025-07-12T05:21:34","guid":{"rendered":"https:\/\/youzum.net\/moonshot-ai-releases-kimi-k2-a-trillion-parameter-moe-model-focused-on-long-context-code-reasoning-and-agentic-behavior\/"},"modified":"2025-07-12T05:21:34","modified_gmt":"2025-07-12T05:21:34","slug":"moonshot-ai-releases-kimi-k2-a-trillion-parameter-moe-model-focused-on-long-context-code-reasoning-and-agentic-behavior","status":"publish","type":"post","link":"https:\/\/youzum.net\/zh\/moonshot-ai-releases-kimi-k2-a-trillion-parameter-moe-model-focused-on-long-context-code-reasoning-and-agentic-behavior\/","title":{"rendered":"Moonshot AI Releases Kimi\u202fK2: A Trillion-Parameter MoE Model Focused on Long Context, Code, Reasoning, and Agentic Behavior"},"content":{"rendered":"<p><strong>Kimi K2<\/strong>, launched by Moonshot AI in July 2025, is a purpose-built, open-source <strong>Mixture-of-Experts (MoE)<\/strong> model\u20141\u202ftrillion total parameters, with <em>32\u202fbillion active parameters<\/em> per token. It\u2019s trained using the custom <strong>MuonClip<\/strong> optimizer on 15.5\u202ftrillion tokens, achieving stable training at this unprecedented scale without the typical instabilities seen in ultra-large models.<\/p>\n<p>Unlike traditional chatbots, K2 is architected specifically for <strong>agentic workflows<\/strong>. It features native <strong>Model Context Protocol (MCP)<\/strong> support and was trained on simulated multi-step tool interactions, enabling it to autonomously decompose tasks, execute tool sequences, write and debug code, analyze data, and orchestrate workflows\u2014all with minimal human oversight.<\/p>\n<h3 class=\"wp-block-heading\"><strong>Why Agentic over Conversational?<\/strong><\/h3>\n<p>While advanced models like GPT-4 and Claude 4 Sonnet excel at language reasoning, <strong>Kimi K2 moves from reasoning to action<\/strong>. It doesn\u2019t just respond\u2014it executes. The core shift lies in enabling real-world workflows:<\/p>\n<ul class=\"wp-block-list\">\n<li><strong>Autonomous code execution<\/strong><\/li>\n<li><strong>Data analysis with charts and interfaces<\/strong><\/li>\n<li><strong>End-to-end web application development<\/strong><\/li>\n<li><strong>Orchestration of 17+ tools per session without human input<\/strong><\/li>\n<\/ul>\n<p>K2\u2019s training incorporated millions of synthetic dialogues, each rated by an LLM-based evaluator. These dialogues simulate realistic tool-use scenarios, giving K2 a practical edge in tool selection and multi-step execution.<\/p>\n<h3 class=\"wp-block-heading\"><strong>Architecture and Training Innovations<\/strong><\/h3>\n<p>K2\u2019s technical design demonstrates several novel elements:<\/p>\n<ul class=\"wp-block-list\">\n<li><strong>MoE Transformer Design<\/strong>: 384 experts with routing to 8 active experts per token, plus 1 shared expert for global context. The model uses 64 attention heads and supports a 128K-token context window.<\/li>\n<li><strong>MuonClip Optimizer<\/strong>: A modified version of Muon that stabilizes training at scale. It uses <strong>qk-clipping<\/strong> to constrain attention scores by rescaling Q\/K matrices, effectively preventing instability in deep layers.<\/li>\n<li><strong>Training Dataset<\/strong>: Over 15.5 trillion tokens from multilingual and multimodal sources, giving K2 robust generalization and tool-use reasoning across diverse domains.<\/li>\n<\/ul>\n<p>The model comes in two variants: <strong>Kimi-K2-Base<\/strong>, the foundational model ideal for fine-tuning and building customized solutions; and <strong>Kimi-K2-Instruct<\/strong>, the post-trained version optimized for immediate use in general-purpose chat and tool-using agentic tasks. Instruct is reflex-grade\u2014optimized for fast, low-latency interaction rather than long-form deliberation. On benchmarks, Kimi K2 outperforms Claude Sonnet 4 and GPT-4.1 in coding and agentic reasoning, with <strong>71.6% on SWE-bench<\/strong>, <strong>65.8% on agentic tasks<\/strong>, and <strong>53.7% on LiveCodeBench<\/strong>.<\/p>\n<h3 class=\"wp-block-heading\"><strong>Performance Benchmarks<\/strong><\/h3>\n<p>Kimi K2 not only matches but often surpasses closed-source models on key benchmarks:<\/p>\n<figure class=\"wp-block-table\">\n<table class=\"has-fixed-layout\">\n<thead>\n<tr>\n<th>Benchmark<\/th>\n<th>Kimi K2<\/th>\n<th>GPT\u20114.1<\/th>\n<th>Claude Sonnet 4<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td>SWE-bench Verified<\/td>\n<td>71.6\u202f%<\/td>\n<td>54.6\u202f%<\/td>\n<td>~72.7\u202f%<\/td>\n<\/tr>\n<tr>\n<td>Agentic Coding (Tau2)<\/td>\n<td>65.8\u202f%<\/td>\n<td>45.2\u202f%<\/td>\n<td>~61\u202f%<\/td>\n<\/tr>\n<tr>\n<td>LiveCodeBench v6 (Pass@1)<\/td>\n<td>53.7\u202f%<\/td>\n<td>44.7\u202f%<\/td>\n<td>47.4\u202f%<\/td>\n<\/tr>\n<tr>\n<td>MATH-500<\/td>\n<td>97.4\u202f%<\/td>\n<td>92.4\u202f%<\/td>\n<td>\u2013<\/td>\n<\/tr>\n<tr>\n<td>MMLU<\/td>\n<td>89.5\u202f%<\/td>\n<td>~90.4\u202f%<\/td>\n<td>~92.9\u202f%<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/figure>\n<p>Its performance in <strong>agentic benchmarks<\/strong> like Tau2 and LiveCodeBench demonstrates its superior capacity to handle multi-step, real-world coding tasks\u2014outperforming many proprietary models.<\/p>\n<h3 class=\"wp-block-heading\"><strong>Cost Efficiency<\/strong><\/h3>\n<p>Perhaps the most disruptive element is pricing:<\/p>\n<ul class=\"wp-block-list\">\n<li><strong>Claude 4 Sonnet<\/strong>: $3 input \/ $15 output per million tokens<\/li>\n<li><strong>Gemini 2.5 Pro<\/strong>: $2.5 input \/ $15 output<\/li>\n<li><strong>Kimi K2<\/strong>: <strong>$0.60 input \/ $2.50 output<\/strong><\/li>\n<\/ul>\n<p>Kimi K2 is roughly <strong>5x cheaper<\/strong> than Claude or Gemini while offering equal or better performance on several metrics. The cost advantage, combined with open access and support for local deployment, positions K2 as an economically viable alternative for developers, enterprises, and research teams.<\/p>\n<h3 class=\"wp-block-heading\"><strong>Strategic Shift: From Thinking to Acting<\/strong><\/h3>\n<p>Kimi K2 marks a pivotal moment in AI\u2019s evolution\u2014from <strong>thinking agents<\/strong> to <strong>acting systems<\/strong>. With native tool-use capabilities and built-in support for multi-agent protocols, it goes far beyond static chat interfaces. It is capable of triggering workflows, making decisions, executing API calls, and delivering tangible outputs autonomously.<\/p>\n<p>Moreover, its release comes at a time when most such capabilities are either locked behind expensive APIs or limited to research labs. K2 is:<\/p>\n<ul class=\"wp-block-list\">\n<li><strong>Open-source<\/strong>, requiring no subscription<\/li>\n<li><strong>Globally accessible<\/strong>, not limited to US-based deployment<\/li>\n<li><strong>Designed for developers<\/strong>, not just end-users<\/li>\n<\/ul>\n<h3 class=\"wp-block-heading\"><strong>Broader Implications<\/strong><\/h3>\n<ol class=\"wp-block-list\">\n<li><strong>Will agentic architecture become the norm?<\/strong> K2\u2019s strong performance on tool use tasks could push proprietary players to rethink their architectures.<\/li>\n<li><strong>Can open-source efforts from Asia compete at global scale?<\/strong> With K2, Moonshot AI joins others like DeepSeek in showing that top-tier performance doesn\u2019t have to originate from Silicon Valley.<\/li>\n<li><strong>What\u2019s next in the agentic evolution?<\/strong> Future models may combine video, robotics, and embodied reasoning to further expand the scope of what agentic AI can accomplish.<\/li>\n<\/ol>\n<h3 class=\"wp-block-heading\"><strong>Conclusion<\/strong><\/h3>\n<p><strong>Kimi K2<\/strong> isn\u2019t just a bigger model\u2014it\u2019s a blueprint for what comes after the reasoning race: <strong>execution-first AI<\/strong>. By combining trillion-parameter scale, low inference costs, and deeply integrated agentic capabilities, Kimi K2 opens the door for AI systems that do more than generate\u2014they build, act, and solve autonomously.<\/p>\n<figure class=\"wp-block-table\">\n<table class=\"has-fixed-layout\">\n<tbody>\n<tr>\n<td>Check out the\u00a0<strong><a href=\"https:\/\/huggingface.co\/collections\/moonshotai\/kimi-k2-6871243b990f2af5ba60617d\" target=\"_blank\" rel=\"noreferrer noopener\">Models on Hugging Face<\/a> and <a href=\"https:\/\/github.com\/MoonshotAI\/Kimi-K2\" target=\"_blank\" rel=\"noreferrer noopener\">GitHub Page<\/a><\/strong>. All credit for this research goes to the researchers of this project. Also,\u00a0feel free to follow us on\u00a0<strong><a href=\"https:\/\/x.com\/intent\/follow?screen_name=marktechpost\" target=\"_blank\" rel=\"noreferrer noopener\"><mark>Twitter<\/mark><\/a><\/strong>, and\u00a0<strong><a href=\"https:\/\/www.youtube.com\/@Marktechpost\">Youtube<\/a><\/strong>\u00a0and don\u2019t forget to join our\u00a0<strong><a href=\"https:\/\/www.reddit.com\/r\/machinelearningnews\/\" target=\"_blank\" rel=\"noreferrer noopener\">100k+ ML SubReddit<\/a><\/strong>\u00a0and Subscribe to\u00a0<strong><a href=\"https:\/\/www.airesearchinsights.com\/subscribe\" target=\"_blank\" rel=\"noreferrer noopener\">our Newsletter<\/a><\/strong>.<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/figure>\n<p>The post <a href=\"https:\/\/www.marktechpost.com\/2025\/07\/11\/moonshot-ai-releases-kimi-k2-a-trillion-parameter-moe-model-focused-on-long-context-code-reasoning-and-agentic-behavior\/\">Moonshot AI Releases Kimi\u202fK2: A Trillion-Parameter MoE Model Focused on Long Context, Code, Reasoning, and Agentic Behavior<\/a> appeared first on <a href=\"https:\/\/www.marktechpost.com\/\">MarkTechPost<\/a>.<\/p>","protected":false},"excerpt":{"rendered":"<p>Kimi K2, launched by Moonshot AI in July 2025, is a purpose-built, open-source Mixture-of-Experts (MoE) model\u20141\u202ftrillion total parameters, with 32\u202fbillion active parameters per token. It\u2019s trained using the custom MuonClip optimizer on 15.5\u202ftrillion tokens, achieving stable training at this unprecedented scale without the typical instabilities seen in ultra-large models. Unlike traditional chatbots, K2 is architected specifically for agentic workflows. It features native Model Context Protocol (MCP) support and was trained on simulated multi-step tool interactions, enabling it to autonomously decompose tasks, execute tool sequences, write and debug code, analyze data, and orchestrate workflows\u2014all with minimal human oversight. Why Agentic over Conversational? While advanced models like GPT-4 and Claude 4 Sonnet excel at language reasoning, Kimi K2 moves from reasoning to action. It doesn\u2019t just respond\u2014it executes. The core shift lies in enabling real-world workflows: Autonomous code execution Data analysis with charts and interfaces End-to-end web application development Orchestration of 17+ tools per session without human input K2\u2019s training incorporated millions of synthetic dialogues, each rated by an LLM-based evaluator. These dialogues simulate realistic tool-use scenarios, giving K2 a practical edge in tool selection and multi-step execution. Architecture and Training Innovations K2\u2019s technical design demonstrates several novel elements: MoE Transformer Design: 384 experts with routing to 8 active experts per token, plus 1 shared expert for global context. The model uses 64 attention heads and supports a 128K-token context window. MuonClip Optimizer: A modified version of Muon that stabilizes training at scale. It uses qk-clipping to constrain attention scores by rescaling Q\/K matrices, effectively preventing instability in deep layers. Training Dataset: Over 15.5 trillion tokens from multilingual and multimodal sources, giving K2 robust generalization and tool-use reasoning across diverse domains. The model comes in two variants: Kimi-K2-Base, the foundational model ideal for fine-tuning and building customized solutions; and Kimi-K2-Instruct, the post-trained version optimized for immediate use in general-purpose chat and tool-using agentic tasks. Instruct is reflex-grade\u2014optimized for fast, low-latency interaction rather than long-form deliberation. On benchmarks, Kimi K2 outperforms Claude Sonnet 4 and GPT-4.1 in coding and agentic reasoning, with 71.6% on SWE-bench, 65.8% on agentic tasks, and 53.7% on LiveCodeBench. Performance Benchmarks Kimi K2 not only matches but often surpasses closed-source models on key benchmarks: Benchmark Kimi K2 GPT\u20114.1 Claude Sonnet 4 SWE-bench Verified 71.6\u202f% 54.6\u202f% ~72.7\u202f% Agentic Coding (Tau2) 65.8\u202f% 45.2\u202f% ~61\u202f% LiveCodeBench v6 (Pass@1) 53.7\u202f% 44.7\u202f% 47.4\u202f% MATH-500 97.4\u202f% 92.4\u202f% \u2013 MMLU 89.5\u202f% ~90.4\u202f% ~92.9\u202f% Its performance in agentic benchmarks like Tau2 and LiveCodeBench demonstrates its superior capacity to handle multi-step, real-world coding tasks\u2014outperforming many proprietary models. Cost Efficiency Perhaps the most disruptive element is pricing: Claude 4 Sonnet: $3 input \/ $15 output per million tokens Gemini 2.5 Pro: $2.5 input \/ $15 output Kimi K2: $0.60 input \/ $2.50 output Kimi K2 is roughly 5x cheaper than Claude or Gemini while offering equal or better performance on several metrics. The cost advantage, combined with open access and support for local deployment, positions K2 as an economically viable alternative for developers, enterprises, and research teams. Strategic Shift: From Thinking to Acting Kimi K2 marks a pivotal moment in AI\u2019s evolution\u2014from thinking agents to acting systems. With native tool-use capabilities and built-in support for multi-agent protocols, it goes far beyond static chat interfaces. It is capable of triggering workflows, making decisions, executing API calls, and delivering tangible outputs autonomously. Moreover, its release comes at a time when most such capabilities are either locked behind expensive APIs or limited to research labs. K2 is: Open-source, requiring no subscription Globally accessible, not limited to US-based deployment Designed for developers, not just end-users Broader Implications Will agentic architecture become the norm? K2\u2019s strong performance on tool use tasks could push proprietary players to rethink their architectures. Can open-source efforts from Asia compete at global scale? With K2, Moonshot AI joins others like DeepSeek in showing that top-tier performance doesn\u2019t have to originate from Silicon Valley. What\u2019s next in the agentic evolution? Future models may combine video, robotics, and embodied reasoning to further expand the scope of what agentic AI can accomplish. Conclusion Kimi K2 isn\u2019t just a bigger model\u2014it\u2019s a blueprint for what comes after the reasoning race: execution-first AI. By combining trillion-parameter scale, low inference costs, and deeply integrated agentic capabilities, Kimi K2 opens the door for AI systems that do more than generate\u2014they build, act, and solve autonomously. Check out the\u00a0Models on Hugging Face and GitHub Page. All credit for this research goes to the researchers of this project. Also,\u00a0feel free to follow us on\u00a0Twitter, and\u00a0Youtube\u00a0and don\u2019t forget to join our\u00a0100k+ ML SubReddit\u00a0and Subscribe to\u00a0our Newsletter. The post Moonshot AI Releases Kimi\u202fK2: A Trillion-Parameter MoE Model Focused on Long Context, Code, Reasoning, and Agentic Behavior appeared first on MarkTechPost.<\/p>","protected":false},"author":2,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"pmpro_default_level":"","site-sidebar-layout":"default","site-content-layout":"","ast-site-content-layout":"","site-content-style":"default","site-sidebar-style":"default","ast-global-header-display":"","ast-banner-title-visibility":"","ast-main-header-display":"","ast-hfb-above-header-display":"","ast-hfb-below-header-display":"","ast-hfb-mobile-header-display":"","site-post-title":"","ast-breadcrumbs-content":"","ast-featured-img":"","footer-sml-layout":"","theme-transparent-header-meta":"","adv-header-id-meta":"","stick-header-meta":"","header-above-stick-meta":"","header-main-stick-meta":"","header-below-stick-meta":"","astra-migrate-meta-layouts":"default","ast-page-background-enabled":"default","ast-page-background-meta":{"desktop":{"background-color":"var(--ast-global-color-4)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"tablet":{"background-color":"","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"mobile":{"background-color":"","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""}},"ast-content-background-meta":{"desktop":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"tablet":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"mobile":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""}},"_pvb_checkbox_block_on_post":false,"footnotes":""},"categories":[52,5,7,1],"tags":[],"class_list":["post-24596","post","type-post","status-publish","format-standard","hentry","category-ai-club","category-committee","category-news","category-uncategorized","pmpro-has-access"],"acf":[],"yoast_head":"<!-- This site is optimized with the Yoast SEO plugin v25.3 - https:\/\/yoast.com\/wordpress\/plugins\/seo\/ -->\n<title>Moonshot AI Releases Kimi\u202fK2: A Trillion-Parameter MoE Model Focused on Long Context, Code, Reasoning, and Agentic Behavior - YouZum<\/title>\n<meta name=\"description\" content=\"\u0e01\u0e34\u0e08\u0e01\u0e23\u0e23\u0e21\u0e40\u0e01\u0e35\u0e48\u0e22\u0e27\u0e01\u0e31\u0e1a\u0e42\u0e14\u0e23\u0e19\" \/>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/youzum.net\/zh\/moonshot-ai-releases-kimi-k2-a-trillion-parameter-moe-model-focused-on-long-context-code-reasoning-and-agentic-behavior\/\" \/>\n<meta property=\"og:locale\" content=\"zh_CN\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"Moonshot AI Releases Kimi\u202fK2: A Trillion-Parameter MoE Model Focused on Long Context, Code, Reasoning, and Agentic Behavior - YouZum\" \/>\n<meta property=\"og:description\" content=\"\u0e01\u0e34\u0e08\u0e01\u0e23\u0e23\u0e21\u0e40\u0e01\u0e35\u0e48\u0e22\u0e27\u0e01\u0e31\u0e1a\u0e42\u0e14\u0e23\u0e19\" \/>\n<meta property=\"og:url\" content=\"https:\/\/youzum.net\/zh\/moonshot-ai-releases-kimi-k2-a-trillion-parameter-moe-model-focused-on-long-context-code-reasoning-and-agentic-behavior\/\" \/>\n<meta property=\"og:site_name\" content=\"YouZum\" \/>\n<meta property=\"article:publisher\" content=\"https:\/\/www.facebook.com\/DroneAssociationTH\/\" \/>\n<meta property=\"article:published_time\" content=\"2025-07-12T05:21:34+00:00\" \/>\n<meta name=\"author\" content=\"admin NU\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:label1\" content=\"\u4f5c\u8005\" \/>\n\t<meta name=\"twitter:data1\" content=\"admin NU\" \/>\n\t<meta name=\"twitter:label2\" content=\"\u9884\u8ba1\u9605\u8bfb\u65f6\u95f4\" \/>\n\t<meta name=\"twitter:data2\" content=\"4 \u5206\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\/\/schema.org\",\"@graph\":[{\"@type\":\"Article\",\"@id\":\"https:\/\/youzum.net\/moonshot-ai-releases-kimi-k2-a-trillion-parameter-moe-model-focused-on-long-context-code-reasoning-and-agentic-behavior\/#article\",\"isPartOf\":{\"@id\":\"https:\/\/youzum.net\/moonshot-ai-releases-kimi-k2-a-trillion-parameter-moe-model-focused-on-long-context-code-reasoning-and-agentic-behavior\/\"},\"author\":{\"name\":\"admin NU\",\"@id\":\"https:\/\/yousum.gpucore.co\/#\/schema\/person\/97fa48242daf3908e4d9a5f26f4a059c\"},\"headline\":\"Moonshot AI Releases Kimi\u202fK2: A Trillion-Parameter MoE Model Focused on Long Context, Code, Reasoning, and Agentic Behavior\",\"datePublished\":\"2025-07-12T05:21:34+00:00\",\"mainEntityOfPage\":{\"@id\":\"https:\/\/youzum.net\/moonshot-ai-releases-kimi-k2-a-trillion-parameter-moe-model-focused-on-long-context-code-reasoning-and-agentic-behavior\/\"},\"wordCount\":790,\"commentCount\":0,\"publisher\":{\"@id\":\"https:\/\/yousum.gpucore.co\/#organization\"},\"articleSection\":[\"AI\",\"Committee\",\"News\",\"Uncategorized\"],\"inLanguage\":\"zh-Hans\",\"potentialAction\":[{\"@type\":\"CommentAction\",\"name\":\"Comment\",\"target\":[\"https:\/\/youzum.net\/moonshot-ai-releases-kimi-k2-a-trillion-parameter-moe-model-focused-on-long-context-code-reasoning-and-agentic-behavior\/#respond\"]}]},{\"@type\":\"WebPage\",\"@id\":\"https:\/\/youzum.net\/moonshot-ai-releases-kimi-k2-a-trillion-parameter-moe-model-focused-on-long-context-code-reasoning-and-agentic-behavior\/\",\"url\":\"https:\/\/youzum.net\/moonshot-ai-releases-kimi-k2-a-trillion-parameter-moe-model-focused-on-long-context-code-reasoning-and-agentic-behavior\/\",\"name\":\"Moonshot AI Releases Kimi\u202fK2: A Trillion-Parameter MoE Model Focused on Long Context, Code, Reasoning, and Agentic Behavior - YouZum\",\"isPartOf\":{\"@id\":\"https:\/\/yousum.gpucore.co\/#website\"},\"datePublished\":\"2025-07-12T05:21:34+00:00\",\"description\":\"\u0e01\u0e34\u0e08\u0e01\u0e23\u0e23\u0e21\u0e40\u0e01\u0e35\u0e48\u0e22\u0e27\u0e01\u0e31\u0e1a\u0e42\u0e14\u0e23\u0e19\",\"breadcrumb\":{\"@id\":\"https:\/\/youzum.net\/moonshot-ai-releases-kimi-k2-a-trillion-parameter-moe-model-focused-on-long-context-code-reasoning-and-agentic-behavior\/#breadcrumb\"},\"inLanguage\":\"zh-Hans\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\/\/youzum.net\/moonshot-ai-releases-kimi-k2-a-trillion-parameter-moe-model-focused-on-long-context-code-reasoning-and-agentic-behavior\/\"]}]},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\/\/youzum.net\/moonshot-ai-releases-kimi-k2-a-trillion-parameter-moe-model-focused-on-long-context-code-reasoning-and-agentic-behavior\/#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Home\",\"item\":\"https:\/\/youzum.net\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"Moonshot AI Releases Kimi\u202fK2: A Trillion-Parameter MoE Model Focused on Long Context, Code, Reasoning, and Agentic Behavior\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\/\/yousum.gpucore.co\/#website\",\"url\":\"https:\/\/yousum.gpucore.co\/\",\"name\":\"YouSum\",\"description\":\"\",\"publisher\":{\"@id\":\"https:\/\/yousum.gpucore.co\/#organization\"},\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\/\/yousum.gpucore.co\/?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"zh-Hans\"},{\"@type\":\"Organization\",\"@id\":\"https:\/\/yousum.gpucore.co\/#organization\",\"name\":\"Drone Association Thailand\",\"url\":\"https:\/\/yousum.gpucore.co\/\",\"logo\":{\"@type\":\"ImageObject\",\"inLanguage\":\"zh-Hans\",\"@id\":\"https:\/\/yousum.gpucore.co\/#\/schema\/logo\/image\/\",\"url\":\"https:\/\/youzum.net\/wp-content\/uploads\/2024\/11\/tranparent-logo.png\",\"contentUrl\":\"https:\/\/youzum.net\/wp-content\/uploads\/2024\/11\/tranparent-logo.png\",\"width\":300,\"height\":300,\"caption\":\"Drone Association Thailand\"},\"image\":{\"@id\":\"https:\/\/yousum.gpucore.co\/#\/schema\/logo\/image\/\"},\"sameAs\":[\"https:\/\/www.facebook.com\/DroneAssociationTH\/\"]},{\"@type\":\"Person\",\"@id\":\"https:\/\/yousum.gpucore.co\/#\/schema\/person\/97fa48242daf3908e4d9a5f26f4a059c\",\"name\":\"admin NU\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"zh-Hans\",\"@id\":\"https:\/\/yousum.gpucore.co\/#\/schema\/person\/image\/\",\"url\":\"https:\/\/youzum.net\/wp-content\/uploads\/avatars\/2\/1746849356-bpfull.png\",\"contentUrl\":\"https:\/\/youzum.net\/wp-content\/uploads\/avatars\/2\/1746849356-bpfull.png\",\"caption\":\"admin NU\"},\"url\":\"https:\/\/youzum.net\/zh\/members\/adminnu\/\"}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"Moonshot AI Releases Kimi\u202fK2: A Trillion-Parameter MoE Model Focused on Long Context, Code, Reasoning, and Agentic Behavior - YouZum","description":"\u0e01\u0e34\u0e08\u0e01\u0e23\u0e23\u0e21\u0e40\u0e01\u0e35\u0e48\u0e22\u0e27\u0e01\u0e31\u0e1a\u0e42\u0e14\u0e23\u0e19","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/youzum.net\/zh\/moonshot-ai-releases-kimi-k2-a-trillion-parameter-moe-model-focused-on-long-context-code-reasoning-and-agentic-behavior\/","og_locale":"zh_CN","og_type":"article","og_title":"Moonshot AI Releases Kimi\u202fK2: A Trillion-Parameter MoE Model Focused on Long Context, Code, Reasoning, and Agentic Behavior - YouZum","og_description":"\u0e01\u0e34\u0e08\u0e01\u0e23\u0e23\u0e21\u0e40\u0e01\u0e35\u0e48\u0e22\u0e27\u0e01\u0e31\u0e1a\u0e42\u0e14\u0e23\u0e19","og_url":"https:\/\/youzum.net\/zh\/moonshot-ai-releases-kimi-k2-a-trillion-parameter-moe-model-focused-on-long-context-code-reasoning-and-agentic-behavior\/","og_site_name":"YouZum","article_publisher":"https:\/\/www.facebook.com\/DroneAssociationTH\/","article_published_time":"2025-07-12T05:21:34+00:00","author":"admin NU","twitter_card":"summary_large_image","twitter_misc":{"\u4f5c\u8005":"admin NU","\u9884\u8ba1\u9605\u8bfb\u65f6\u95f4":"4 \u5206"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"Article","@id":"https:\/\/youzum.net\/moonshot-ai-releases-kimi-k2-a-trillion-parameter-moe-model-focused-on-long-context-code-reasoning-and-agentic-behavior\/#article","isPartOf":{"@id":"https:\/\/youzum.net\/moonshot-ai-releases-kimi-k2-a-trillion-parameter-moe-model-focused-on-long-context-code-reasoning-and-agentic-behavior\/"},"author":{"name":"admin NU","@id":"https:\/\/yousum.gpucore.co\/#\/schema\/person\/97fa48242daf3908e4d9a5f26f4a059c"},"headline":"Moonshot AI Releases Kimi\u202fK2: A Trillion-Parameter MoE Model Focused on Long Context, Code, Reasoning, and Agentic Behavior","datePublished":"2025-07-12T05:21:34+00:00","mainEntityOfPage":{"@id":"https:\/\/youzum.net\/moonshot-ai-releases-kimi-k2-a-trillion-parameter-moe-model-focused-on-long-context-code-reasoning-and-agentic-behavior\/"},"wordCount":790,"commentCount":0,"publisher":{"@id":"https:\/\/yousum.gpucore.co\/#organization"},"articleSection":["AI","Committee","News","Uncategorized"],"inLanguage":"zh-Hans","potentialAction":[{"@type":"CommentAction","name":"Comment","target":["https:\/\/youzum.net\/moonshot-ai-releases-kimi-k2-a-trillion-parameter-moe-model-focused-on-long-context-code-reasoning-and-agentic-behavior\/#respond"]}]},{"@type":"WebPage","@id":"https:\/\/youzum.net\/moonshot-ai-releases-kimi-k2-a-trillion-parameter-moe-model-focused-on-long-context-code-reasoning-and-agentic-behavior\/","url":"https:\/\/youzum.net\/moonshot-ai-releases-kimi-k2-a-trillion-parameter-moe-model-focused-on-long-context-code-reasoning-and-agentic-behavior\/","name":"Moonshot AI Releases Kimi\u202fK2: A Trillion-Parameter MoE Model Focused on Long Context, Code, Reasoning, and Agentic Behavior - YouZum","isPartOf":{"@id":"https:\/\/yousum.gpucore.co\/#website"},"datePublished":"2025-07-12T05:21:34+00:00","description":"\u0e01\u0e34\u0e08\u0e01\u0e23\u0e23\u0e21\u0e40\u0e01\u0e35\u0e48\u0e22\u0e27\u0e01\u0e31\u0e1a\u0e42\u0e14\u0e23\u0e19","breadcrumb":{"@id":"https:\/\/youzum.net\/moonshot-ai-releases-kimi-k2-a-trillion-parameter-moe-model-focused-on-long-context-code-reasoning-and-agentic-behavior\/#breadcrumb"},"inLanguage":"zh-Hans","potentialAction":[{"@type":"ReadAction","target":["https:\/\/youzum.net\/moonshot-ai-releases-kimi-k2-a-trillion-parameter-moe-model-focused-on-long-context-code-reasoning-and-agentic-behavior\/"]}]},{"@type":"BreadcrumbList","@id":"https:\/\/youzum.net\/moonshot-ai-releases-kimi-k2-a-trillion-parameter-moe-model-focused-on-long-context-code-reasoning-and-agentic-behavior\/#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https:\/\/youzum.net\/"},{"@type":"ListItem","position":2,"name":"Moonshot AI Releases Kimi\u202fK2: A Trillion-Parameter MoE Model Focused on Long Context, Code, Reasoning, and Agentic Behavior"}]},{"@type":"WebSite","@id":"https:\/\/yousum.gpucore.co\/#website","url":"https:\/\/yousum.gpucore.co\/","name":"YouSum","description":"","publisher":{"@id":"https:\/\/yousum.gpucore.co\/#organization"},"potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/yousum.gpucore.co\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"zh-Hans"},{"@type":"Organization","@id":"https:\/\/yousum.gpucore.co\/#organization","name":"Drone Association Thailand","url":"https:\/\/yousum.gpucore.co\/","logo":{"@type":"ImageObject","inLanguage":"zh-Hans","@id":"https:\/\/yousum.gpucore.co\/#\/schema\/logo\/image\/","url":"https:\/\/youzum.net\/wp-content\/uploads\/2024\/11\/tranparent-logo.png","contentUrl":"https:\/\/youzum.net\/wp-content\/uploads\/2024\/11\/tranparent-logo.png","width":300,"height":300,"caption":"Drone Association Thailand"},"image":{"@id":"https:\/\/yousum.gpucore.co\/#\/schema\/logo\/image\/"},"sameAs":["https:\/\/www.facebook.com\/DroneAssociationTH\/"]},{"@type":"Person","@id":"https:\/\/yousum.gpucore.co\/#\/schema\/person\/97fa48242daf3908e4d9a5f26f4a059c","name":"admin NU","image":{"@type":"ImageObject","inLanguage":"zh-Hans","@id":"https:\/\/yousum.gpucore.co\/#\/schema\/person\/image\/","url":"https:\/\/youzum.net\/wp-content\/uploads\/avatars\/2\/1746849356-bpfull.png","contentUrl":"https:\/\/youzum.net\/wp-content\/uploads\/avatars\/2\/1746849356-bpfull.png","caption":"admin NU"},"url":"https:\/\/youzum.net\/zh\/members\/adminnu\/"}]}},"rttpg_featured_image_url":null,"rttpg_author":{"display_name":"admin NU","author_link":"https:\/\/youzum.net\/zh\/members\/adminnu\/"},"rttpg_comment":0,"rttpg_category":"<a href=\"https:\/\/youzum.net\/zh\/category\/ai-club\/\" rel=\"category tag\">AI<\/a> <a href=\"https:\/\/youzum.net\/zh\/category\/committee\/\" rel=\"category tag\">Committee<\/a> <a href=\"https:\/\/youzum.net\/zh\/category\/news\/\" rel=\"category tag\">News<\/a> <a href=\"https:\/\/youzum.net\/zh\/category\/uncategorized\/\" rel=\"category tag\">Uncategorized<\/a>","rttpg_excerpt":"Kimi K2, launched by Moonshot AI in July 2025, is a purpose-built, open-source Mixture-of-Experts (MoE) model\u20141\u202ftrillion total parameters, with 32\u202fbillion active parameters per token. It\u2019s trained using the custom MuonClip optimizer on 15.5\u202ftrillion tokens, achieving stable training at this unprecedented scale without the typical instabilities seen in ultra-large models. Unlike traditional chatbots, K2 is architected&hellip;","_links":{"self":[{"href":"https:\/\/youzum.net\/zh\/wp-json\/wp\/v2\/posts\/24596","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/youzum.net\/zh\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/youzum.net\/zh\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/youzum.net\/zh\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/youzum.net\/zh\/wp-json\/wp\/v2\/comments?post=24596"}],"version-history":[{"count":0,"href":"https:\/\/youzum.net\/zh\/wp-json\/wp\/v2\/posts\/24596\/revisions"}],"wp:attachment":[{"href":"https:\/\/youzum.net\/zh\/wp-json\/wp\/v2\/media?parent=24596"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/youzum.net\/zh\/wp-json\/wp\/v2\/categories?post=24596"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/youzum.net\/zh\/wp-json\/wp\/v2\/tags?post=24596"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}