Z.ai says GLM-5.3 leads CyberGym with 84.5% and found 2,436 vulnerabilities. Independent verifications: zero. Vulnerabilities with public CVEs: 53 out of 2,436.

CAPABILITY Z.AI (LAUNCH POST, AUG 14) · CLAIMED 2026-08-1408/13

Z.ai says GLM-5.3 leads CyberGym with 84.5% and found 2,436 vulnerabilities. Independent verifications: zero. Vulnerabilities with public CVEs: 53 out of 2,436.

Every number is self-reported by the company that made the model. The weights are withheld for two weeks, making independent verification impossible until August 28.

01THE CLAIM

"GLM-5.3 leads CyberGym at 84.5%, beating Anthropic Mythos 5 and GPT-5.6 Sol, and found 2,436 confirmed vulnerabilities; weights delayed two weeks for safety." [SOURCE ↗]

BS4 SOURCES · LIVE 2026-08-25

Z.AI TRACK RECORD1 CLAIM · 100/100 BS RATE →

84.5% CyberGym score per OfficeChai

0independent verifications at launch

2436claimed confirmed vulnerabilities per fello AI

0.7point margin over Mythos 5, derived: 84.5 minus 83.8, both from Z.ai's own chart per OfficeChai receipt

02THE CHECK

THE CLAIM. Z.ai launched GLM-5.3 on August 14 claiming CyberGym 84.5%, beating Mythos 5 (83.8%) and GPT-5.6 Sol (83.6%), plus 2,436 confirmed vulnerabilities across 269 open-source projects.

THE CHECK. searched for independent CyberGym replication of GLM-5.3 on August 16, 2026: zero results. Every score is Z.ai-reported, Z.ai-tested, on Z.ai's own harness configuration. Of the 2,436 "confirmed" vulnerabilities, only 53 have public CVE assignments. The remaining 2,383 are under embargo. Confirmed is doing work it has not earned.

THE TWIST. the weights are withheld for two weeks "for safety testing." That doubles as a perfect scarcity launch window. No one can replicate the benchmark until after the press cycle ends.

03SAY THIS IN THE MEETING · 📸 SCREENSHOT IT

"GLM-5.3 claims CyberGym first place by 0.7 points on its own test. Independent verifications: zero. The weights are held back for two weeks. That is not a benchmark result. That is a press embargo dressed as safety."

Z.ai launched GLM-5.3 on August 14 with a familiar playbook: announce a model, cite a benchmark, claim the top spot, and withhold the weights that would let anyone check.

🔒 THE FULL AUTOPSY · FREE WITH AN ACCOUNT

You just read the free check. Sign in free, a code by email, no passwords, and the rest unlocks: the evidence trail, the steelman and the rebuttal, all 4 sources with quotes and screenshots, and our on-record call.