GLM-5.3: Post-training-only gains push open-weight coding and exploit capability to the top

GLM-5.3: Frontier Coding with Emergent Cyber Capabilities

Z.ai released GLM-5.3 with the same base model as 5.2 — every gain is from post-training. Coding jumped 50% on their internal Z.ai Code Bench, and Terminal Bench 3.0 went from 4.6 to 28.3. The bigger surprise: exploit capability grew far faster than expected. ExploitGym 2h score rose from 29 to 105, 6h from 39 to 130. The team credits training environments that mirror real expert workflows, pushing the model to chain full exploit sequences. Weights will be open-sourced in two weeks after safety hardening.

Why it matters: Zhipu releases GLM-5.3 — same base model as 5.2, all gains from post-training. Code bench up 50%, Terminal Bench from 4.6 to 28.3, 2-hour exploit score from 29 to 105. The lab admits cyber capability emerged faster than expected. Domestic flagship model launch with concrete nu...