Chinese AI Model GLM-5.2 Rivals Top US Models Amid Export Restrictions
Beijing-based Zhipu AI (Z.ai) released GLM-5.2, an open-source large language model trained on Huawei Ascend chips, challenging US AI dominance. The model nearly matches Anthropic's Opus 4.8 on agentic benchmarks at one-fifth the cost, topping leaderboards after the US banned Anthropic's Fable 5 model. This event intensifies the US-China AI competition, highlighting open-source advantages and China's ability to innovate despite chip export controls.
Editorial summary awaiting refresh
Cross-source coverage
Wire timeline
Zhipu AI's GLM-5.2 Challenges US AI Dominance in Coding and Cybersecurity
Chinese AI company Zhipu AI (Z.ai) has unveiled GLM-5.2, its most advanced large language model, on June 16, 2026. The model, released as an open-weight system, is reported to seriously challenge leading US systems like Anthropic's Claude Opus 4.8 in specialized coding and cybersecurity tasks. Tests by cybersecurity firm Semgrep showed GLM-5.2 outperforming Anthropic's models on certain software engineering and bug-finding tasks. The launch, accompanied by the release of the ZCode AI coding platform, drew comparisons to DeepSeek's earlier breakthrough and has been called a 'DeepSeek moment' for China's open-source AI ecosystem. The development comes amid tightened US export controls on advanced chips and AI technologies, as China intensifies home-grown innovation. GLM-5.2 has entered the top 10 most-used AI systems on OpenRouter, highlighting its growing developer adoption.
The Hindu: Latest News today from India and the World, Breaking news, Top Headlines and Trending News Videos.Chinese Z.ai's GLM-5.2 tops AI leaderboards after US bans Anthropic Fable 5, using Huawei chips
On June 12, the US Commerce Department banned Anthropic from supplying its Fable 5 or Mythos 5 AI models to foreign nationals, forcing a global shutdown. The next day, Beijing-based Z.ai (formerly Zhipu AI) released GLM-5.2 under an MIT license, trained entirely on Huawei Ascend 910B chips without Nvidia hardware. Within a week, GLM-5.2 topped several AI leaderboards, including Design Arena's human-preference coding board and Artificial Analysis's Intelligence Index, surpassing Fable 5 in some benchmarks. It scored 62.1 on SWE-bench Pro vs GPT-5.5's 58.6. However, it trailed in multi-week knowledge tasks and terminal work. The export controls designed to hinder Chinese AI development appear to have failed, as Z.ai, on the US Entity List since January 2025, produced a competitive model using domestic silicon. Anthropic contested the ban, calling it based on a misunderstanding, but complied. GLM-5.2's massive 744-billion-parameter model requires high-end hardware to run.
Latest from Tom's HardwareChinese AI Firm Z.ai's GLM-5.2 Tops Rankings After U.S. Ban on Anthropic's Fable 5
Following a U.S. Commerce Department directive on June 12th barring Anthropic from supplying its Fable 5 and Mythos 5 models to foreign nationals, Beijing-based Z.ai (formerly Zhipu AI) released GLM-5.2, an open-weight AI model trained entirely on Huawei Ascend 910B chips with no Nvidia silicon. Within a week, GLM-5.2 climbed to the top of openly available AI leaderboards, surpassing Anthropic's Fable 5 on coding benchmarks and ranking first on Artificial Analysis's Intelligence Index. Z.ai's market value passed HK$1 trillion. The model demonstrates that frontier-class open models can be produced using a fully domestic Chinese chip stack, despite U.S. export controls aimed at curbing China's AI development. However, analyses note hardware parity remains elusive, with Huawei's Ascend 910C at about 60% of Nvidia H100 inference performance. Anthropic disabled Fable 5 globally, citing a narrow jailbreak vulnerability, and disputed the Commerce Department's rationale.
Latest from Tom's HardwareChinese Z.ai's Open-Weight GLM-5.2 Tops AI Leaderboards After US Ban on Anthropic Fable 5
Following a June 12th US Commerce Department directive barring Anthropic from supplying its Fable 5 model to foreign nationals, Beijing-based Z.ai (formerly Zhipu AI) launched GLM-5.2, an open-weight AI model trained entirely on Huawei Ascend chips. Within a week, GLM-5.2 climbed to the top of openly available AI leaderboards, achieving first place on Design Arena's coding board and Artificial Analysis's Intelligence Index v4.1. Z.ai's market value surpassed HK$1 trillion (US$128 billion). The model's training on 100,000 Huawei Ascend 910B processors, with no Nvidia hardware, challenges US export controls designed to curtail Chinese AI development. However, GLM-5.2 lags on longer-term and terminal benchmarks behind Anthropic's now-disabled models. Experts note the achievement demonstrates model capability parity with domestic chips but not hardware parity with Nvidia, as US chips remain significantly more powerful.
Latest from Tom's HardwareChina's Zhipu AI Model GLM 5.2 Nears Top US Rivals as Cost and Open Source Shift Dynamics
Chinese AI startup Zhipu has released GLM 5.2, an open-source model that scores within one percentage point of Anthropic's Opus 4.8 on a key agentic benchmark at roughly a fifth of the cost. The model excels in planning, coding, and testing, attracting enterprise interest as companies seek better value per dollar on token spending. Developer traffic on OpenRouter spiked faster than after DeepSeek's earlier launch. The release gains additional traction as the Trump administration restricts US frontier models, forcing Anthropic to pull its Fable model and OpenAI to limit GPT-5.6 to trusted partners. This combination of high performance, low cost, and immunity from government revocation makes open source a compelling alternative for enterprises.
US Top News and AnalysisChina's Zhipu GLM 5.2 Nears Top US AI Models as Open Source Gains Ground
Chinese AI startup Zhipu launched its open source GLM 5.2 model, which nearly matches Anthropic's Opus 4.8 on a key agentic benchmark at roughly a fifth of the cost. The model has attracted significant developer interest, with token traffic on OpenRouter growing faster than after DeepSeek's V4 launch. GLM 5.2 excels in agentic tasks like planning, coding, and testing, which enterprises are eager to automate. Amid rising token costs, the metric of intelligence per dollar is becoming paramount. Additionally, US government restrictions on Anthropic's Fable and OpenAI's GPT-5.6 have made open source models, which cannot be revoked, an attractive alternative for enterprises seeking reliable access to frontier AI.
US Top News and AnalysisChinese AI Model GLM-5.2 Stirs Excitement in Silicon Valley
The Chinese AI company z.AI has released GLM-5.2, an open-source large language model designed for long-form coding tasks and agentic workflows. The model features a 1 million token context window, placing it in competition with leading US models like Anthropic's Claude Opus 4.8 and OpenAI's GPT 5.5. The model has garnered significant attention and praise from Silicon Valley executives, including Vercel CEO Guillermo Rauch and former Meta VP Matt Velloso, who called it the first open model suitable for daily use. The release reignites debate over whether the US is maintaining its AI lead against China, especially as open-source models like GLM-5.2 and DeepSeek's R1 challenge the dominance of closed American systems. The article also notes US concerns about China closing the gap via chip restrictions and distillation attacks.
All Content from Business Insider