Zhipu Releases GLM-5.3: Post-Training Scaling Boosts Coding & Security, Open Source in Two Weeks
智谱 · wechat · 2026-08-14
Zhipu has released GLM-5.3, which leverages extreme post-training scaling on an unchanged base to significantly boost coding, agentic, and cybersecurity capabilities. The model ranks first among open-source models in benchmarks like TerminalBench and approaches closed-source frontiers in coding experience, while offering notably better token efficiency.
The model has demonstrated emergent cyber defense capabilities. Partnering with security teams, GLM-5.3 uncovered over a thousand medium-to-high-risk vulnerabilities across OSs, DNS protocols, and humanoid robots. Emphasizing that 'the best shield must belong to everyone,' Zhipu announced the model will be open-sourced in two weeks and launched the 'Open Source Shield' initiative to provide free security auditing capabilities to the community.
More from Models
- New inference engine 'e' hits 10,500 tok/s on a single DGX Station — brandon_galang · 2026-10-16
- NVIDIA's Nemotron 3.5 Super VL hits CoreWeave: 12B-active multimodal MoE at $0.25/M input tokens — wandb · 2026-10-16
- Opus 5 might be a very deep model or use looping, commentator speculates — scaling01 · 2026-10-16
- GPT-6.1 Sol audit finds 6% of arXiv papers had false main theorems, down to 1.5% recently — nitarshan · 2026-10-16
- Nvidia-backed post-training pushes Nemotron 3.5 VL past frontier on dermatology diagnosis — LoganGrasby · 2026-10-16
- Gemini 4 Argon scores 97.5% on ARC-AGI-3, second lab above 95% — patience_cave · 2026-10-16