GLM-5.3 ships on same base as 5.2: post-training alone lifts Terminal Bench 4.6→28.3
burny_tech · x · 2026-09-08
Zhipu has launched the GLM-5.3 API, targeting coding, defensive cybersecurity, and long-horizon agentic tasks, priced the same as GLM-5.2. Researcher QGallouedec highlights the key detail: GLM-5.3 uses the identical base model as 5.2 with no pretraining changes — post-training (RL) alone lifted Terminal Bench 3.0 from 4.6 to 28.3 and made it #1 on CyberGym, beating all closed frontier models. A strong datapoint that "the frontier is RL."
More from Models
- Model slips on a science knowledge test, author calls it acceptable — felpix_ · 2026-09-08
- Native video inference beats sampled frames, but OpenRouter providers don't support it yet — spillai · 2026-09-08
- Users game Tibo's token reset: spend 80% fast, stagger reset cycles to dodge surprise resets — tinyfool · 2026-09-08
- Economists in the top 10% of AI use see no step change from GPT 5.6 to 6 — aniketapanjwani · 2026-09-08
- Mathematician: Astra solved two of my unpublished theorems in ~60 hours each, $20k in API — basedjensen · 2026-09-08
- Drop-in Claude.md patch claims to fix Claude's degraded prose in Opus and Fable — myriable · 2026-09-08