Open-source GLM-5.3 nearly matches Claude in exploit capability

Andrew Ng's latest The Batch highlights Anthropic's evaluation showing open-weights GLM-5.3 scoring 12% vs Claude's 14% on ExploitBench exploitation tasks, and finding a Chrome vulnerability for around $20.

2026-10-02 ~ 2026-10-04 · 2 related posts

• The Batch: Open-Weights GLM-5.3 Nears Claude on Vulnerability Exploitation, 12% vs 14% — DeepLearningAI · 2026-10-02
• GLM-5.3 nearly matches Claude at exploiting bugs; $20 in tokens found a Chrome flaw — DeepLearningAI · 2026-10-04