Chinese developer Zhipu AI says its open-weight GLM-5.2 model now matches Anthropic's Mythos at finding software security flaws.
Key Points:
- GLM-5.2, built by Beijing's Zhipu AI, now rivals Mythos at spotting software vulnerabilities, researchers say.
- The open-weight model still trails Anthropic and OpenAI on broader reasoning work.
- Its rise pressures Washington as officials tighten export limits on top U.S. AI systems.
GLM-5.2 Matches Mythos On Bugs
The Wall Street Journal reported that researchers found GLM-5.2 even with Mythos at scanning code for vulnerabilities in some scenarios, even as it trails American leaders on most other tasks.
Zhipu AI, also known as Z.ai, released the model on Jun. 13 under a permissive open license. Anyone can download the system and run it on ordinary hardware, and it already ranks among the ten most-used models on one large routing service.
Independent benchmarks lend further weight to the claim.
In tests by the security firm Semgrep, GLM-5.2 outscored Anthropic's Claude Opus 4.8 on several vulnerability checks, and researchers say added prompting lifted both systems to Mythos's level at spotting exploitable bugs.
Also Read: Chainlink's Wallet Record Turns LINK's $9 Rebound Into The Main Test
Why China's AI Surge Worries Washington
Chinese rivals are pressing the advantage. Security firm 360 Security Technology rolled out a bug-hunting tool called Tulongfeng, and its chief, Zhou Hongyi, argued that such firepower cannot remain solely in American hands.
Microsoft and other platforms are now weighing whether to host Chinese models, a move that could reshape how businesses buy AI.
The shift lands at an awkward moment for Washington, where officials treat these systems as prized national-security assets in the AI race. Anthropic only recently regained limited U.S. clearance for Mythos after a temporary freeze, while the Trump administration pressed it and OpenAI to restrict foreign access and stagger releases such as GPT-5.6.
Critics call the strategy self-defeating. Use of Chinese systems has surged as firms chase lower costs, and security researcher Niels Provos warned that locking up American models only pushes more users toward open-weight rivals.
Saif Khan, who helped write earlier U.S. export rules, said curbing sales while China builds its own tools rewards Beijing and weakens U.S. cyber defenses.
Mythos Set The Cyber Bar
The public debate over how fast that gap would close had already spilled into open view. Tech figure Elon Musk recently predicted Chinese labs would match Anthropic's flagship on benchmarks by early 2027, and Zhipu founder Tang Jie shot back that real progress would not take that long.
Mythos built its fearsome reputation in a matter of weeks. Under Anthropic's Project Glasswing, the model uncovered a 27-year-old flaw in OpenBSD, helped Mozilla find 271 bugs in an early Firefox build, and surfaced thousands more flaws across widely used open-source code. Those results turned the system into the yardstick that every rival now openly races to match.
Read Next: XRP Falls Near $1 While ETF Buyers Test A Weak Spot Market