Z.ai says GLM-5.3 scores 84.5% on CyberGym, vs. Mythos 5's 83.8%, and its most sensitive cybersecurity functions will only be available to verified users (Reuters)
Z.ai is framing GLM-5.3 as a near-Mythos cyber model, but with staged access and verification gates.
The Techmeme item says Z.ai’s open-source GLM-5.3 reached 84.5% on CyberGym, narrowly above Mythos 5’s 83.8%. Reuters describes the model as nearing Anthropic’s restricted Mythos 5 in software vulnerability identification. Z.ai says the most sensitive cybersecurity functions will be limited to verified users. API access and open weights are set to roll out in stages after safety evaluations. Techmeme's note
The Techmeme item says Z.ai’s open-source GLM-5.3 reached 84.5% on CyberGym, narrowly above Mythos 5’s 83.8%. Reuters describes the model as nearing Anthropic’s restricted Mythos 5 in software vulnerability identification. Z.ai says the most sensitive cybersecurity functions will be limited to verified users. API access and open weights are set to roll out in stages after safety evaluations. Techmeme's note
score 7