--- title: GLM-5.2: DeepSeek Was Wrong About RL? | daily.dev description: GLM 5.2 from ZhipuAI is positioned as a major open-weight agentic model release under MIT license, ranking among the top models globally at launch while being... canonical: https://daily.dev/posts/glm-5-2-deepseek-was-wrong-about-rl--0i3c5ngt6 twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: GLM-5.2: DeepSeek Was Wrong About RL? | daily.dev og:description: GLM 5.2 from ZhipuAI is positioned as a major open-weight agentic model release under MIT license, ranking among the top models globally at launch while being... og:url: https://daily.dev/posts/glm-5-2-deepseek-was-wrong-about-rl--0i3c5ngt6 og:image: https://api.daily.dev/og/posts/0i3C5nGt6.png og:image:alt: GLM-5.2: DeepSeek Was Wrong About RL? og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # GLM-5.2: DeepSeek Was Wrong About RL? **[bycloud](https://daily.dev/sources/bycloud)** · 15 min read · 0 upvotes · 0 comments ## Summary GLM 5.2 from ZhipuAI is positioned as a major open-weight agentic model release under MIT license, ranking among the top models globally at launch while being significantly cheaper than Claude and GPT-5. The key technical insight is ZhipuAI's decision to abandon GRPO (DeepSeek's group relative policy optimization) in favor of PPO with a critic model for RL training, arguing that GRPO's group-comparison approach breaks down for long-horizon agentic tasks with variable-length trajectories. They built a training infrastructure called SLIME to standardize diverse agent trajectories across environments. For long-context efficiency, they introduced 'index share' — having every four sparse attention layers share one indexer — reducing indexing compute by 75% and achieving ~2.9x fewer per-token FLOPs at 1M context. The model also uses parallel on-policy distillation merging 10+ expert models. The post also discusses the open-source business model tradeoffs for ZhipuAI and broader implications for the AI inference ecosystem. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Similar posts on daily.dev - [GLM-5.2: Built for Long-Horizon Tasks](https://daily.dev/posts/glm-5-2-built-for-long-horizon-tasks-stmnmwjok) · Hugging Face · 33 upvotes · 4 comments - [A Bit Late but Ultimate Analysis: DeepSeek](https://daily.dev/posts/a-bit-late-but-ultimate-analysis-deepseek-cvg0jqn04) · Serokell · 1 upvotes · 0 comments --- Tags: [#open-source](https://daily.dev/tags/open-source), [#llm](https://daily.dev/tags/llm), [#ai-agents](https://daily.dev/tags/ai-agents), [#reinforcement-learning](https://daily.dev/tags/reinforcement-learning) [View this post on daily.dev](https://daily.dev/posts/glm-5-2-deepseek-was-wrong-about-rl--0i3c5ngt6) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"GLM-5.2: DeepSeek Was Wrong About RL?","url":"https://daily.dev/posts/glm-5-2-deepseek-was-wrong-about-rl--0i3c5ngt6","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/glm-5-2-deepseek-was-wrong-about-rl--0i3c5ngt6"},"datePublished":"2026-07-23T21:24:36.627Z","dateModified":"2026-07-23T21:25:33.791Z","description":"GLM 5.2 from ZhipuAI is positioned as a major open-weight agentic model release under MIT license, ranking among the top models globally at launch while being...","image":"https://i.ytimg.com/vi/3KwpmSpEplY/sddefault.jpg","thumbnailUrl":"https://i.ytimg.com/vi/3KwpmSpEplY/sddefault.jpg","isAccessibleForFree":true,"articleSection":"bycloud","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"bycloud","logo":"https://media.daily.dev/image/upload/s--UUgAibgb--/f_auto/v1712315122/logos/bycloud","url":"https://daily.dev/sources/bycloud"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/glm-5-2-deepseek-was-wrong-about-rl--0i3c5ngt6","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"open-source,llm,ai-agents,reinforcement-learning","timeRequired":"PT15M","video":{"@type":"VideoObject","name":"GLM-5.2: DeepSeek Was Wrong About RL?","description":"GLM 5.2 from ZhipuAI is positioned as a major open-weight agentic model release under MIT license, ranking among the top models globally at launch while being...","thumbnailUrl":"https://i.ytimg.com/vi/3KwpmSpEplY/sddefault.jpg","uploadDate":"2026-07-23T21:24:36.627Z","duration":"PT15M","url":"https://api.daily.dev/r/0i3C5nGt6","embedUrl":"https://www.youtube.com/embed/3KwpmSpEplY"}} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"bycloud","item":"https://daily.dev/sources/bycloud"},{"@type":"ListItem","position":3,"name":"GLM-5.2: DeepSeek Was Wrong About RL?"}]} ```