China's GLM-5.2 matches Mythos on bugs — and it's open weights
zhipu ai's open-weight glm-5.2 reportedly matches the proprietary mythos model on bug-finding and cybersecurity tasks, narrowing the capability gap between chinese and us models despite export controls.
zhipu ai launched glm-5.2, an open-weight large language model that benchmarks suggest performs on par with anthropic's mythos in identifying software vulnerabilities and cybersecurity tasks. the release contradicts the effectiveness of current us export controls designed to stifle china's ai progress by limiting access to advanced chips and top-tier model weights. unlike closed us models, glm-5.2 can be downloaded and run locally on consumer-grade hardware, offering flexibility but also reducing barriers to potential misuse by malicious actors. us security agencies view advanced cybersecurity models as critical national security threats, making this localized parity a strategic challenge.