Anthropic says Zhipu's open-weight GLM-5.3 nearly matches Claude Mythos Preview at building exploits
Anthropic flags Zhipu's open GLM-5.3: nearly matches Claude Mythos at writing exploits, costs $20.40 per attack.
Why it matters
Anthropic publicly warns that Zhipu's open-weight GLM-5.3 matches or approaches Claude's capability for malicious code generation — a capability-parity claim backed by CAISI, with real exploit-cost data and evidence of stripped safeguards already in circulation.
The key facts
5 to knowZhipu GLM-5.3 open-weight model writes cyber exploits nearly as well as Claude Mythos Preview
GLM-5 Flash variant generated reliable Chrome attack for $20.40 at Zhipu API pricing
Model safeguards described as easy to strip; unlocked versions already circulating
Anthropic claim backed by US agency CAISI
Open-weight model with measurable exploit-generation capability parity to frontier closed model
Go to the source
The Decoderthe-decoder.com
Publisher excerpt: Zhipu's open-weight model GLM-5.3 writes cyber exploits nearly as well as Claude Mythos Preview, according to Anthropic. Its smaller Flash variant put together a reliable Chrome attack for just $20.40 at Zhipu's API prices. The model's safeguards are easy to strip out, and unlocked versions are…