Z.ai, also known as Zhipu, has released GLM-5.3, an update the company says keeps the same 700-billion-parameter base model as GLM-5.2 and derives every capability gain from scaled-up post-training. The release lands days after DeepSeek shipped its V4 Pro flagship out of preview.
Stronger coding through environment scaling
GLM-5.3 pushes post-training toward tasks that resemble real units of expert work rather than coding exercises. The company built pipelines that synthesize environments end to end: research agents convert task patterns from real work into runnable long-horizon environments, a judge agent verifies solvability, and verifiers are synthesized without access to reference solutions.
Results show the largest gains on the longest-horizon evaluations. At High effort, GLM-5.3 reaches 31.4 percent at roughly 50,000 output tokens per task, surpassing Claude Opus 4.8 at 29.5 percent with 120,000 tokens. It remains behind Claude Fable 5 at 39.5 percent at maximum effort.
| Benchmark | GLM-5.3 | GLM-5.2 | Change |
|---|---|---|---|
| Terminal-Bench 3.0 | 28.3% | 4.6% | +23.7pp |
| DeepSWE v1.1 | 66.9% | 46.2% | +20.7pp |
| SWE-Marathon v1.1 | 42.5% | 19.4% | +23.1pp |
| Z.ai Code Bench (Max effort) | 34.5% | 23.4% | +11.1pp |
The model also achieves open-source state of the art on public benchmarks including Terminal Bench 3.0 and Agents' Last Exam.
Emergent cyber capability
The second headline is one Z.ai flags as unexpected. The company introduced vulnerability discovery data and environments into post-training, expecting better isolated bug-finding. Instead, capability continued compounding as training scaled, and the model began reasoning across multiple stages of exploitation, forming coherent plans for complete exploitation chains.
| Benchmark | GLM-5.3 | GLM-5.2 | Mythos 5 | GPT-5.6 Sol |
|---|---|---|---|---|
| CyberGym | 84.5% | 77.2% | 83.8% | 83.6% |
| ExploitBench | 54.4% | 24.4% | 78.0% | 76.5% |
| ExploitGym (2h / 6h) | 105 / 130 | 29 / 39 | 181 / 247 | 216 / 293 |
The pattern is consistent: the further up the exploitation chain a benchmark sits, the larger the gain from GLM-5.2, and the wider the remaining gap to closed frontier models. Z.ai reports testing transfer beyond controlled benchmarks. Working with several security teams in China, the company says its models have identified 2,436 vulnerabilities across 269 open-source projects since GLM-5.2, including 1,097 rated critical or high severity, spanning system kernels, operating systems, browser engines, and network protocols.
Safety and release timeline
The two-week gap between announcement and weight release is tied explicitly to safety evaluation and hardening. Z.ai says the weights will be downloadable after this period. GLM-5.3 is available now through Z.ai's API and its GLM Coding Plan, rolled out to all existing subscribers.
For the API, GLM-5.3 supports three thinking effort levels (low, high, max) and no longer permits disabling thinking, a breaking change for applications that previously ran with thinking switched off.
Sources
Z.ai, GLM-5.3: Frontier Coding with Emergent Cyber Capabilities (Aug 14, 2026): https://z.ai/blog/glm-5.3
Unite AI, Z.ai Launches GLM-5.3 With Frontier Coding and a Cyber Capability That Outgrew Its Training (Aug 14, 2026): https://www.unite.ai/z-ai-launches-glm-5-3-with-frontier-coding-and-a-cyber-capability-that-outgrew-its-training/
Bloomberg, Z.ai Aims to Catch Anthropic, OpenAI in Coding With New AI Model (Aug 14, 2026): https://www.bloomberg.com/news/articles/2026-08-14/z-ai-aims-to-catch-anthropic-openai-in-coding-with-new-ai-model
Z.ai Security Disclosure Ledger: https://cvd.z.ai/
