CAISI evaluates Zhipu GLM-5.2 on cyber benchmarks
terryyuezhuo · x · 2026-07-20
CAISI’s evaluation of Zhipu’s GLM-5.2 is highlighted as a quiet but notable update amid the Kimi news cycle.
- The shared charts compare cyber capability across models on ExploitBench and CTF Archive Diamond.
- GLM-5.2 is shown against models such as Mythos Preview, Opus 4.8, GPT-5.5, and DeepSeek V4 Pro.
- On the ExploitBench chart, GLM-5.2 is labeled with a 21% success rate, while GPT-5.5 is shown at 41% and Mythos Preview at 57%.
- On CTF Archive Diamond, GLM-5.2 is shown at 39%, with GPT-5.5 at 71% and DeepSeek V4 Pro at 32%.
- A second chart suggests a large jump from SFT to SFT→RL on seen and unseen compositions, indicating substantial gains from RL after supervised tuning.
More from Models
- Nat Lambert: Cohere's model is narrow RL training on internal infra ill-suited for open weights — natolambert · 2026-10-27
- OpenAI teases DOTS is coming to Europe 'soon' — mark_k · 2026-10-27
- Clifford Pickover challenges AI to decipher a message he encoded in Norse runes — pickover · 2026-10-27
- ChatGPT's crawler sample code strips out all Schema markup — VeryWellVersed · 2026-10-27
- Cohere sandbox: Qwen model tried to upload probe file to Beijing cloud server unprompted — simonguozirui · 2026-10-27
- Kimi K3.1 Pro tops open-weight models on FrontierML benchmark, still far behind Opus 5.5 — teortaxesTex · 2026-10-27