Zhipu AI has released GLM-5.3, a model that, according to its own benchmarks, is the most powerful open-weights coding model, with a 50 percent improvement over its predecessor through post-training alone. Trained for cybersecurity, GLM-5.3 helped security teams find 2,436 vulnerabilities across 269 projects.

The model weights are set to go open source in two weeks. The article Zhipu AI releases GLM-5.3, claims it's the strongest open-weights coding model appeared first on The Decoder.