GATE86The China Gateway
Z.ai's GLM-5.3 matches top US models on elite hacking benchmarks

Z.ai's GLM-5.3 matches top US models on elite hacking benchmarks

China's Z.ai (also known as Zhipu AI) delayed the public release of GLM-5.3 after discovering its vulnerability reproduction success rate reached 84.5% — fractionally above Anthropic's Claude at 83.8% on the same benchmark.

Subscriber Content

Subscribe to read the full analysis

From $9/month · 7-day free trial

Already a subscriber? Log in

Read more from Industry →

Subscribe for full access to all articles

Start Free Trial