The Interconnects "Latest open artifacts #24" newsletter (September 8, 2026) highlights a notable licensing change for GLM-5.3: Z.ai switched from the MIT license used by GLM-5.2 and earlier releases to a custom license. The new terms add a Model-as-a-Service clause requiring any licensee, together with its affiliates, The newsletter flags that "affiliates" is not defined in the license, creating adoption uncertainty, and notes the license text is provided in both English and Chinese with the Chinese using 关联方 for affiliated parties, which has a defined meaning under Chinese law. While the USD 10 billion revenue threshold is descr

Z.ai announced GLM-5.3 on August 14, 2026, positioning it as an open-weight model built for advanced coding and cybersecurity work that shares its base architecture with GLM 5.2, with every capability gain attributed to expanded post-training rather than a larger base. The company also launched OpenVuln, a scanning ser According to the BetaNews report, GLM-5.3 scored 84.5% on CyberGym and 54.4% on ExploitBench in Z.ai's own testing (up from 77.2% and 24.4% for GLM-5.2), and is credited with 2,436 vulnerability findings across 269 open-source projects, including 1,097 rated critical or high severity. Z.ai said it would withhold GLM-5.

Zhipu AI launched GLM-5.3 on August 14, 2026 as its next-generation foundation model with 743 billion parameters, with most performance gains coming from post-training rather than base-model scaling. According to Zhipu's technical report, GLM-5.3 scored 28.3 on Terminal-Bench 3.0, which the company called the highest r Zhipu positioned GLM-5.3 as the strongest open-source model across several mainstream benchmarks, claiming its coding and agent capabilities approach those of Anthropic's Claude Fable 5 (likely a translation/OCR rendering of a Claude model name), with practical coding performance surpassing other Chinese models. The la

GLM-5.3 was announced on August 14, 2026 by Z.ai (Zhipu AI) as a post-training-only release built on the same base model as GLM-5.2, with every performance gain attributed to post-training according to Z.ai's blog and documentation. The model is text-only, always-on reasoning with a 1M-token context window, 128K max ou Working with security teams in China, Z.ai reports that GLM-5.3 surfaced 2,436 vulnerabilities across 269 projects, including 1,097 medium-to-high severity issues with an average latent lifetime of 26.6 years and the oldest flaw introduced in 1981. The architecture discussion covers GLM-5.2's IndexShare change, improve