# GLM-5.3 scores 12% on ExploitBench, near Claude Mythos at 14%

DeepLearning.AI’s weekly letter cites Anthropic figures showing GLM-5.3 solving 12% of ExploitBench exploit tasks and Claude Mythos 14%. It also says $20.40 in tokens found a recent Google Chrome security exploit.

Language: en
Time zone: America/Los_Angeles
Current time (UTC, cached as of): 2026-11-22T12:52:00.000Z
Query date (America/Los_Angeles): 2026-11-22

HTML: https://didcodexreset.com/news/e0988a83f234446ca49012c3.html

Content language: en
Localization state: sameLanguage

Source: DeepLearning.AI · Published 10/3/2026, 20:21:06

[DeepLearning.AI’s weekly letter](https://hubs.la/Q04z3MJH0) says an open model, GLM-5.3, nearly matches Claude Mythos on cybersecurity tasks. According to Anthropic, GLM-5.3 solves 12% of ExploitBench exploit tasks and Claude Mythos solves 14%.

The letter says $20.40 in tokens found a recent security exploit in Google Chrome. Andrew Ng writes that concern about these capabilities in the wrong hands should be treated as an engineering problem, and that defenders hold the long-term edge.

Tags: GLM-5.3, Claude Mythos, ExploitBench, Cybersecurity, Andrew Ng

[View original post](https://x.com/DeepLearningAI/status/2106580062619255053)
