Reading up on Z.ai
3 deep · digging since sep 15
- Open-weight AI models can exploit the patch gap
Open-weight model GLM 5.3 exploited the Firefox patch gap, turning public security fixes into one-click exploits for six releases at about $1,000 each.
- Toward Recursive Self-Improvement: How GLM Built Its Own Inference Infrastructure
GLM developed a production inference service on over 100,000 domestically made AI accelerators to run its GLM-5.3‑Flash model and achieved significant performance gains through aggressive memory optimizations.
- Mouse on GLM-5.3-Flash
Mouse's open‑source harness achieved a 77% pass rate on FrontierHarness using GLM‑5.3‑Flash at $0.29 per run, close to top performers but far cheaper.