Zhipu Ai · v5.3 · depuis 14. August 2026 · 5× · vu le 15 août 2026
31
Momentum
GLM-5.3 is Zhipu AI's (Z.ai) current flagship language model, released on August 14, 2026. It uses the same 743B-parameter MoE base as its predecessor GLM-5.2, with all improvements coming exclusively from extended post-training. It focuses on software engineering, agentic tasks, and cybersecurity (including vulnerability discovery); Z.ai reports open-source state-of-the-art results on benchmarks such as Terminal-Bench 3.0, Agents' Last Exam, and CyberGym. Open model weights are delayed (about two weeks after launch) for safety reasons; a separate per-token API price had not yet been published at the time of this research.
Historique du momentum
17.05.15.08.
Fonctionnalités
| Key Benchmark (%) | Terminal-Bench 3.0: increased from 4.6 to 28.3; Agents' Last Exam: increased from 23.8 to 28.5 |
| Context Window (Tokens) | 1,000,000 tokens (base model identical to GLM-5.2) |
| License | Open weights planned but delayed ~2 weeks post-launch for safety review (not yet on Hugging Face) |
| Multimodality | Text-focused (coding/agentic); no confirmed vision input in GLM-5.3, though requested by the community |
| Platform | Available via GLM Coding Plan and ZCode; Z.ai API access staged/coming soon |
| Price per 1M Tokens | Not yet published separately; GLM Coding Plan starts at $18/month (Lite), $80 (Pro), $168 (Max) |
| Release Date | August 14, 2026 |