Z AI's latest model update focuses on top tier coding and agentic capabilities achieved through post training on a 743B base model. The GLM-5.3 Max variant scored 1,597 points in the Code Arena: WebDev benchmark, placing it eighth overall and second among open weight options. At $3.65 per million tokens, the model outperforms proprietary variants including Deepseek-v4-flash-high and Gemini 3.7 Flash-high.

The GLM-5.3 API is currently available via first party and partner gateways for tasks involving defensive cybersecurity and long horizon agentic work. Z AI expects to release the model weights shortly to support the open ecosystem, shifting the performance and cost balance for developers.

Sign in to suggest edits

Key sources

  1. SOURCE@arena“At $3.65/M, GLM-5.3 (Max) is on par with Qwen3.8 (Max) ($5.00/M)”x.com
  2. SUPPORT@zai_org“shifted the Pareto frontier for Code Arena: WebDev!”x.com
  3. SUPPORT@scaling01“GLM-5.3 is crushing Terminal-Bench 3.0”x.com
Markdown