Наталя ХандусенкоAI Eng
14 August 2026, 11:41
2026-08-14
China's Z.ai announced its competitor to Anthropic and OpenAI in AI coding — GLM-5.3
Beijing-based startup Z.ai is releasing a new version of its flagship AI model, GLM-5.3. The main focus is on improving code writing and analysis skills, which should bring it closer to American leaders, in particular Anthropic's Fable 5 model.
Beijing-based startup Z.ai is releasing a new version of its flagship AI model, GLM-5.3. The main focus is on improving code writing and analysis skills, which should bring it closer to American leaders, in particular Anthropic's Fable 5 model.
The new model is built on the same database of approximately 700 billion parameters as the previous version, which was released only in June, writes Bloomberg.
Z.ai plans to release the model's weights within two weeks. Published benchmarks show that GLM-5.3 significantly outperforms the successful GLM-5.2 and in many tests breathes on the back (and sometimes even beats) Fable 5.
“Compared to GLM-5.2, it copes much better with complex coding and long-running tasks,” the company’s blog says .
Introducing GLM-5.3: Built to Code. Ready for Cyber Defense.
— Top-tier coding and agent capabilities, achieved through post-training on the 743B base model — A major leap in cybersecurity, setting a new standard among open models
GLM-5.3, an open-weights programming model, performed 50% better than GLM-5.2 on the internal Z.ai Code Bench benchmark. It also achieved open-source SOTA (best-of-breed open-source solution) performance in public benchmarks, including Terminal Bench 3.0 and Agents' Last Exam.
The company admitted that the model learned cybersecurity even faster than planned. It became the best in tests for finding vulnerabilities in systems. But the most interesting thing is that it learned not only to find these “weak spots”, but also several times better at creating tools for their actual hacking.
Z.ai is one of the pioneers of Chinese AI and the world’s first large-scale language model developer to go public. At its peak after the release of GLM-5.2, the company’s market value soared to $137 billion, overtaking internet giants like PDD Holdings and NetEase. It has since fallen to around $80 billion, still a tenfold increase since its Hong Kong listing in January.
Як нейромережі бачать вільну та незалежну Україну? Тест dev.ua
Нейронні мережі для генерації зображень бачать світ по-своєму, їхню логіку зрозуміти часом зовсім неможливо. Але таки хочеться. На честь Дня Незалежності України редакція dev.ua вирішила провести невеликий експеримент.
Ми задали чотирьом різним нейронним мережам п’ять однакових запитів: «прапор України», «День Незалежності України», «український Крим», «перемога України» та «українці». Отриманими результатами ми ділимося з вами нижче.
У TikTok тепер можна генерувати фон за допомогою нейромережі. Ми протестували її та ділимося результатами
У TikTok з’явилася нова функція «Розумний фон». З її допомогою як фон для тіктоків можна підставляти згенеровані нейромережею зображення. Редакція dev.ua протестувала цю технологію і ділиться своїми враженнями.