China's Z.ai releases GLM-5.3, rivaling Mythos 5 in cyber-security benchmarks
Chinese firm Z.ai has announced that its latest open-source model, GLM-5.3, is nearly as effective as Anthropic’s restricted Mythos 5 at spotting software security flaws. The company stated that GLM-5.3 achieved a score of 84.5% on the CyberGym platform, which evaluates cybersecurity performance, slightly edging out the 83.8% score attributed to Mythos 5. However, the model did lag behind Mythos 5 when it came to turning identified vulnerabilities into functional exploits, a key aspect of defensive security testing.
by shortkt.com✓