Zhipu's GLM-5.3 scores 84.5% on CyberGym, edging out Anthropic and OpenAI models
pstAsiatech · x · 2026-08-14
Zhipu (Z.ai) has unveiled its flagship GLM-5.3 model, claiming it beats US frontier systems on a key cybersecurity benchmark, SCMP reports:
- CyberGym (tests whether models can identify and validate security flaws from source code): GLM-5.3 hit 84.5%, above Anthropic's Mythos 5 at 83.8% and OpenAI's GPT-5.6 Sol at 83.6%.
- ExploitBench (gauges how far models climb the exploitation ladder): GLM-5.3's 54.4% trailed Mythos at 78% and GPT-5.6 Sol at 76.5%.
- Zhipu says it tested the model with security teams in China on real-world codebases, identifying 2,436 vulnerabilities across 269 projects after expert review, 1,097 rated medium-to-high severity.
The poster also flags a broader concern: open-weight models plus pre-release testing.
More from Models
- Google says teams across the company are using Gemini 4 Argon on its hardest engineering problems — CurieuxExplorer · 2026-10-15
- Perplexity's new embedding models: index with a 9B model, search with 0.6B — antoine_chaffin · 2026-10-15
- 797GB of n-gram counts in CPU RAM let 1B models match 57% more compute — PangWeiKoh · 2026-10-15
- Leak: Moonshot's Kimi flagship 'x4' done with pre-training, K3.1 in internal testing — lyraxana · 2026-10-15
- Unsloth ships GGUF quants for Nemotron 3.5: Bolt-74B runs in 24GB RAM — danielhanchen · 2026-10-15
- NVIDIA ships two open Nemotron 3.5 Bolt models: a 74B-A9B MoE and a mobile-friendly E4B — HankYeomans · 2026-10-15