Zhipu · v5.3 · 10× · tolest 15. Aug. 2026
GLM-5.3 is a language model released by Zhipu AI (Z.ai) on August 14, 2026, built on the same 743-billion-parameter base model as GLM-5.2. Performance improvements stem exclusively from intensified Post-Training (more training environments, extended RL phase), not from a new base architecture. According to Zhipu, the model achieves top scores among Open Source models on Coding, Agenten, and Cybersecurity Benchmarks (including Terminal-Bench 3.0 and CyberGym), but does not yet reach the level of leading closed models like Claude Opus/Mythos or GPT-5.6 Sol on complex Exploitation tasks. Open weights are scheduled to be released two weeks after launch; a standalone API price list had not yet been published at the time of research.
Features
| Key Benchmark (%) | CyberGym: 84.5% (surpassing Mythos 5 at 83.8% and GPT-5.6 Sol at 83.6%) |
| Context Window (Tokens) | 1,000,000 tokens (1M context, per Coding Plan documentation glm-5.3[1m]) |
| License | Open weights planned two weeks after release (open-weights, license details not yet officially confirmed) |
| Multimodality | Text-only; no native vision support confirmed (GLM-5.3 launch post makes no mention of vision) |
| Platform | Available via GLM Coding Plan (Z.ai); standalone API listed as "coming soon" |
| Price per 1M Tokens | Not yet officially published; GLM Coding Plan starts at $18/month (entry), $80 (Pro), $168 (Max) |
| Release Date | August 14, 2026 |