{"rewrite":{"id":"r_26fa319be611646f04b2c298","clusterId":"c_013040e6debd83eb2031e5ef","slug":"open-weight-model-glm-5-2-beats-claude-code-in-vulnerability-detection-test","model":"deepseek-v4-flash","headline":"Open-Weight Model GLM-5.2 Beats Claude Code in Vulnerability Detection Test","summary":"Chinese AI company Z.ai's open-weight model GLM-5.2 outperformed Anthropic's Claude Code in a benchmark measuring detection accuracy for Insecure Direct Object Reference vulnerabilities, according to a report from cybersecurity firm Semgrep. GLM-5.2 placed third overall, ahead of multiple Claude Code configurations, with a cost per vulnerability found of about $0.17.","whyItMatters":"An open-weight model with simple prompts beat a state-of-the-art coding agent on a reasoning-intensive security task, and at a fraction of the cost.","webCardHtml":"\u003cp\u003eSemgrep tested AI models including GLM-5.2, Claude Code (Opus 4.6/4.7/4.8), and GPT-5.5 on a benchmark for Insecure Direct Object Reference detection. GLM-5.2, a mixture-of-experts model with approximately 750 billion parameters, scored 39% F1, 7 percentage points above Claude Code\u0026#39;s 32%. Semgrep called the result the biggest surprise of the test. GLM-5.2 also cost about $0.17 per vulnerability found, roughly one-sixth the cost of comparable cutting-edge models. Z.ai released the model on June 17, and founder Jie Tang referenced the suspension of Claude Fable\u0026#39;s service due to a U.S. government order in the announcement.\u003c/p\u003e","blueskyPost":"Open-weight model GLM-5.2 beat Claude Code in a vulnerability detection benchmark, scoring 39% F1 vs 32%, at $0.17 per find.","twitterPost":"Open-weight model GLM-5.2 beat Claude Code in a vulnerability detection benchmark, scoring 39% F1 vs 32%, at $0.17 per find.","threadsPost":null,"newsletterBlurb":"Z.ai's open-weight model GLM-5.2 outperformed Claude Code in a Semgrep benchmark for detecting Insecure Direct Object Reference vulnerabilities. The model scored 39% F1, 7 points above Claude Code, and cost about $0.17 per vulnerability found.","attributionJson":"[{\"source\":\"GIGAZINE\",\"url\":\"https://gigazine.net/news/20260630-glm-5-2-beats-claude-cyber-benchmarks/\",\"title\":\"China's open-weight model 'GLM-5.2' outperforms Claude Code in vulnerability detection benchmark\"}]","lintFlagsJson":null,"lintHits":0,"costUsd":0,"inputTokens":4326,"outputTokens":565,"status":"published","repairAttempts":0,"nextRepairAt":null,"factsAttemptedAt":1782830450,"createdAt":"2026-06-30T14:35:41.000Z","publishedAt":"2026-06-30T14:39:50.000Z","updatedAt":"2026-06-30T14:39:50.000Z"},"cluster":{"id":"c_013040e6debd83eb2031e5ef","canonicalTitle":"中国のオープンウェイトモデル「GLM-5.2」が脆弱性検出ベンチマークでClaude Codeを上回る","representativeArticleId":"a_61c32bc691cdee4aff677066","sourceCount":2,"writtenSourceCount":1,"writeAttempts":0,"isSolo":false,"entitiesJson":"{\"anime_titles\":[],\"manga_titles\":[],\"work_titles\":[],\"studios\":[],\"people\":[],\"type\":\"news\",\"domain\":\"other\",\"is_roundup\":false}","contentType":"news","status":"published","firstSeenAt":"2026-06-30T14:00:00.000Z","lastSeenAt":"2026-07-02T03:04:44.000Z","updatedAt":"2026-07-02T06:05:45.000Z"},"attribution":[{"source":"ASCII.jp","url":"https://ascii.jp/elem/000/004/415/4415766/?rss","title":"Stella AI、Anthropic社の最新モデル「Claude Sonnet 5」「Claude Sonnet 4.5」を追加"},{"source":"GIGAZINE","url":"https://gigazine.net/news/20260701-claude-sonnet-5/","title":"「Claude Sonnet 5」が登場、Opus 4.8に近い性能で安価＆無料プランでも使用可能"}],"entities":{"anime_titles":[],"manga_titles":[],"work_titles":[],"studios":[],"people":[],"type":"news","domain":"other","is_roundup":false},"keyFacts":["The model costs about $0.17 per vulnerability found, roughly one-sixth the cost of comparable cutting-edge models.","GLM-5.2 is a mixture-of-experts model with approximately 750 billion parameters.","Z.ai released GLM-5.2 on June 17, and founder Jie Tang referenced the suspension of Claude Fable's service due to a U.S. government order in the announcement."]}