Z.ai Ships GLM-5.3 Without Retraining the Base Model: Better at Complex Coding and Long-Horizon Tasks
Z.ai released GLM-5.3 on August 14, 2026. The model reuses the 743B GLM-5.2 base unchanged. Every reported gain comes from scaled post-training: more long-horizon task environments, more environment types, longer training. Terminal-Bench 3.0 moves from 4.6 to 28.3, and DeepSWE v1.1 from 46.2 to 66.9. Cybersecurity moved further than Z.ai says it planned, with CyberGym at 84.5% and ExploitBench more than doubling to 54.4%. Weights arrive in about two weeks. The post Z.ai Ships GLM-5.3 Without Retraining the Base Model: Better at Complex Coding and Long-Horizon Tasks appeared first on MarkTechPo
Put this to work
Try the tools in this story inside the AI Tool Lab, browse roles in the AI Job Portal, test your skills with Aptitude Tests, or read our in-depth guides.
Keep reading
More AI news
Y Combinator’s Garry Tan wants U.S. open-weight AI labs to ‘distill’ frontier models, too
Y Combinator leader Garry Tan is advocating for a shift in how leading artificial intelligence developers share their technology. He contends that since foundational models are built upon vast repositories of human-generated information, the resulting capabilities should be accessible as a public benefit. By encouraging labs to create smaller, distilled versions of frontier models, Tan hopes to democratize access to high-performance AI, moving away from closed-off ecosystems toward a more equitable distribution of innovative tools.
Read the briefOpenAI’s feud with mathematicians is only escalating
A significant group of prominent mathematicians has issued a formal complaint regarding the practices of major AI organizations. They assert that current data scraping and model training methods infringe upon their intellectual property and undermine the integrity of academic research. As AI firms continue to aggressively consume mathematical publications for training data, this conflict highlights the growing tension between rapid technological advancement and the fundamental protections required for scholarly work.
Read the briefLawyer fined $5K over AI-hallucinated witnesses in a murder case
The New Mexico Supreme Court has sanctioned an attorney, Stephen Aarons, following his submission of fabricated evidence in a legal appeal. The attorney utilized generative AI to create non-existent witnesses and police reports, failing to verify the accuracy of the documents before filing them in a murder case. This incident underscores the severe professional risks associated with relying on artificial intelligence tools without rigorous fact-checking and oversight in high-stakes legal environments.
Read the brief