{"rewrite":{"id":"r_d923129f9d88b09b6f4fc6df","clusterId":"c_1bb4a88106f09903869eafd3","slug":"openrouter-sees-surge-in-glm-5-2-api-usage-after-launch","model":"deepseek-v4-flash","headline":"OpenRouter Sees Surge in GLM-5.2 API Usage After Launch","summary":"Chinese AI model GLM-5.2, released by Z.ai on June 17, 2026, has seen rapid adoption on OpenRouter, accounting for 75% of Z.ai's API share within a week. The model outperformed Claude Opus 4.7 and GPT-5.5 in multiple benchmarks and surpassed Claude Fable 5 in coding tests. OpenRouter noted it is spreading faster than most open models, with top use cases including AI agents like Hermes Agent and Cline.","whyItMatters":"GLM-5.2's rapid uptake on OpenRouter signals that an open Chinese model is competing directly with top proprietary models on cost and quality in real-world agent use.","webCardHtml":"\u003cp\u003eZ.ai\u0026#39;s GLM-5.2, released June 17, has become the dominant model in Z.ai\u0026#39;s OpenRouter share at 75% within a week, according to OpenRouter data. The model beat Claude Opus 4.7 and GPT-5.5 on multiple benchmarks and topped Claude Fable 5 in the Design Arena blind coding test. OpenRouter called its spread very rapid for an open model. The most common use case is the AI agent Hermes Agent, followed by agents including pi, Claude Code, Kilo Code, and Cursor. The AI coding agent Cline reported that GLM-5.2 fixed real bugs in Cline at half the cost of Claude Opus 4.8, though it consumed twice the tokens, and that GLM-5.2 verified code compilation before task completion while Claude Opus 4.8 did not.\u003c/p\u003e","blueskyPost":"GLM-5.2's OpenRouter adoption hit 75% of Z.ai's API share in a week. The model's coding benchmark lead over Claude Fable 5 is driving agent use cases like Hermes Agent and Cline.","twitterPost":"GLM-5.2 hit 75% of Z.ai's OpenRouter share in a week. Its coding lead over Claude Fable 5 is fueling agent adoption.","threadsPost":"GLM-5.2 captured 75% of Z.ai's OpenRouter API share within a week of release. The model outperformed Claude Opus 4.7 and GPT-5.5 on benchmarks, and its coding lead over Claude Fable 5 is driving use in AI agents like Hermes Agent and Cline.","newsletterBlurb":"OpenRouter data shows GLM-5.2, a Chinese open model from Z.ai, captured 75% of Z.ai's API share within a week of its June 17 launch. The model outperformed Claude Opus 4.7 and GPT-5.5 on multiple benchmarks and surpassed Claude Fable 5 in a blind coding test. Its top use is in AI agents like Hermes Agent and Cline, with Cline reporting lower cost and better verification than Claude Opus 4.8.","attributionJson":"[{\"source\":\"GIGAZINE\",\"url\":\"https://gigazine.net/news/20260623-glm-5-2-openrouter/\",\"title\":\"Surge in OpenRouter API Users for Chinese Model 'GLM-5.2': What Are AI Users Using It For?\"}]","lintFlagsJson":null,"lintHits":0,"costUsd":0,"inputTokens":4446,"outputTokens":742,"status":"published","repairAttempts":0,"nextRepairAt":null,"factsAttemptedAt":1782207867,"createdAt":"2026-06-23T09:36:39.000Z","publishedAt":"2026-06-23T09:40:24.000Z","updatedAt":"2026-06-23T09:40:24.000Z"},"cluster":{"id":"c_1bb4a88106f09903869eafd3","canonicalTitle":"OpenRouterで中華モデル「GLM-5.2」のAPI利用者が急増、AIユーザーたちは何に使っているのか？","representativeArticleId":"a_0e6942cac268318f1c22b318","sourceCount":1,"writtenSourceCount":1,"writeAttempts":0,"isSolo":true,"entitiesJson":"{\"anime_titles\":[],\"manga_titles\":[],\"work_titles\":[],\"studios\":[],\"people\":[],\"type\":\"news\",\"domain\":\"other\",\"is_roundup\":false}","contentType":"news","status":"published","firstSeenAt":"2026-06-23T09:02:00.000Z","lastSeenAt":"2026-06-23T09:02:00.000Z","updatedAt":"2026-06-23T09:40:24.000Z"},"attribution":[{"source":"GIGAZINE","url":"https://gigazine.net/news/20260623-glm-5-2-openrouter/","title":"OpenRouterで中華モデル「GLM-5.2」のAPI利用者が急増、AIユーザーたちは何に使っているのか？"}],"entities":{"anime_titles":[],"manga_titles":[],"work_titles":[],"studios":[],"people":[],"type":"news","domain":"other","is_roundup":false},"keyFacts":["GLM-5.2, released by Z.ai on June 17, 2026, accounted for 75% of Z.ai's API share on OpenRouter within a week.","The model outperformed Claude Opus 4.7 and GPT-5.5 on multiple benchmarks and surpassed Claude Fable 5 in the Design Arena blind coding test.","OpenRouter noted that GLM-5.2's spread is very rapid for an open model, with top use cases including AI agents like Hermes Agent and Cline."]}