GLM 5.3 Release Date & What's New: Complete Timeline
Quick answer: GLM 5.3 was released on August 14, 2026 by Z.AI — available immediately via the GLM Coding Plan and ZCode, with the public API rolling out soon and open weights on HuggingFace ~2 weeks after launch (expected late August). It's the same base model as GLM 5.2 with all gains from post-training: +50% coding, open-weights SOTA on Terminal-Bench 3.0, and emergent cyber capability.
TL;DR
| Milestone | Date / Window |
|---|---|
| Official release | August 14, 2026 |
| Coding Plan access | August 14, 2026 (all subscribers upgraded) |
| ZCode access | August 14, 2026 |
| Public API | "Coming soon" (after launch) |
| Open weights (HuggingFace) | ~2 weeks after launch (late August) |
| ZCode quota boost ends | August 31, 2026 |
Release Timeline
August 14, 2026 — Launch Day
Z.AI released GLM 5.3 with a blog post titled "GLM-5.3: Frontier Coding with Emergent Cyber Capabilities." The headline facts:
- Same base model as GLM 5.2 — every gain from scaled post-training.
- +50% coding capability on Z.AI's internal code bench.
- Open-weights SOTA on Terminal-Bench 3.0 (28.3 vs 5.2's 4.6) and Agents' Last Exam.
- CyberGym 84.5 — best public vulnerability-discovery result, ahead of GPT-5.6 Sol (83.6).
- 2,436 real-world vulnerabilities found across 269 projects during testing.
Available immediately: GLM Coding Plan (all subscribers auto-upgraded) and ZCode. The API was announced as coming soon, with endpoints pre-published (OpenAI Chat Completion / Response, Anthropic Message protocols).
August 14 – ~August 28 — API Rollout Window
The glm-5.3 API is expected to go live in the days/weeks after launch. Watch for:
- Model ID:
glm-5.3 - Thinking always enabled:
reasoning_effort(low/high/max, defaultmax) - 1M context, 128K max output
- Tool calling with streaming parameter output
OpenRouter and other routers typically add the model within days of the provider API going live — as of launch day, OpenRouter lists z-ai/glm-5.2 but not yet glm-5.3.
~Late August — Open Weights on HuggingFace
Z.AI's stated plan: weights two weeks after launch, once safety evaluation and hardening are complete. The delay is deliberate — GLM 5.3's cyber capability (vulnerability discovery + exploitation planning) is dual-use, so Z.AI is gating the open release on safety work. The HuggingFace page is already marked "Coming Soon."
August 31 — ZCode Boost Deadline
ZCode's 1.5× limited-time quota boost ends August 31 (stacks with 98%+ cache savings for up to 180% standard quota). If you're evaluating GLM 5.3 via ZCode, the cheap stress-test window closes at month end.
What's New vs GLM 5.2 (One Place)
| Area | What changed |
|---|---|
| Coding | +50% on Z.AI Code Bench; TB 3.0 4.6 → 28.3; DeepSWE 46.2 → 66.9 |
| Agents | AutomationBench 26.2 → 48.2; ALE-CLI 23.8 → 28.5 |
| Cyber (emergent) | CyberGym 77.2 → 84.5 (SOTA); ExploitBench 24.4 → 54.4 |
| Efficiency | 34.5% @ ~75K tokens vs 5.2's 23.4% @ ~96K (Max effort) |
| Thinking | Always on — reasoning_effort low/high/max; disabled removed |
| Token efficiency | Better results with fewer output tokens |
| Weights | Public (5.2) → held ~2 weeks for safety (5.3) |
How to Get It Now
- Coding Plan — z.ai/subscribe: Lite $18/mo, Pro $80/mo, Max $168/mo; off-peak hours cost 50% of standard points. Already subscribed? You're on 5.3.
- ZCode — zcode.z.ai: agent-native access with Goal mode and the 1.5× boost through Aug 31.
- API — coming soon; model ID
glm-5.3, thinking always enabled. - Self-host — weights in ~2 weeks; same serving stack as GLM 5.2.
FAQ
When was GLM 5.3 released? August 14, 2026 — via GLM Coding Plan and ZCode immediately; API and open weights follow.
When will GLM 5.3 weights be released? Approximately two weeks after launch (expected late August), pending safety evaluation and hardening.
Is GLM 5.3 available on the API? Not yet — Z.AI says the API is coming soon. Coding Plan and ZCode are the current access paths.
What's the biggest change in GLM 5.3? The same base model as 5.2 with dramatically better post-training: +50% coding, Terminal-Bench 3.0 4.6 → 28.3, and emergent cyber capability (CyberGym SOTA).
When does the ZCode 1.5× boost end? August 31, 2026.
Sources
- Z.AI: GLM-5.3: Frontier Coding with Emergent Cyber Capabilities (August 14, 2026)
- Z.AI Docs: GLM-5.3
Last updated: August 14, 2026




