China AI Hub
Models / GLM-5.3
GLM-5.3
GLM-5 family · Provider: zhipu-ai · Status: active · Released: 2026-08-18
GLM-5.3 is a GLM-5 model developed by zhipu-ai, a Chinese AI company headquartered in Beijing Zhipu Huazhang Technology Co., Ltd. (北京智谱华章科技股份有限公司), Beijing, China, released 2026-08-18 with a 1,048,576-token context window and open weights.
| Key fact | Value |
|---|---|
| Model ID | glm-5.3 |
| Architecture | 744B total / 40B active (open-weight FP8); same base model as GLM-5.2 with post-training gains |
| Context window | 1,048,576 tokens |
| Max output | 131,072 tokens |
| Open weights | Yes |
| License | Apache-2.0 (per GitHub repo metadata; README has no separate weights-license section - verify per-model HF cards before reuse) |
| Self-hosting | Yes |
| API available | Yes |
| API pricing | $1.4 input / $4.4 output per 1M tokens (USD) · provider pricing page |
| Regions | international, china |
| Cloud providers | Z.ai, BigModel |
Capabilities
| Capability | Supported |
|---|---|
| Reasoning | Yes |
| Coding | Yes |
| Math | Unknown |
| Chinese | Unknown |
| English | Unknown |
| Multilingual | Unknown |
| Vision | No |
| Audio | Unknown |
| Video | Unknown |
| Tool calling | Unknown |
| Function calling | Unknown |
| Structured output | Unknown |
| Agent capability | Unknown |
| RAG | Unknown |
| Computer use | Unknown |
Benchmark results
| Benchmark | Version | Score | Metric | Date | Source type | Source |
|---|---|---|---|---|---|---|
| Terminal-Bench 3.0 | — | 28.3 | accuracy | 2026-08-18 | vendor_reported | link |
| DeepSWE v1.1 | — | 66.9 | accuracy | 2026-08-18 | vendor_reported | link |
| Agents' Last Exam (CLI) | — | 28.5 | accuracy | 2026-08-18 | vendor_reported | link |
| CyberGym | — | 84.5 | accuracy | 2026-08-18 | vendor_reported | link |
Benchmark scores are single data points, not universal rankings. Vendor-reported scores are labeled as such.
Known limitations
- Text-only input
- Reasoning always enabled (low/high/max, default max); cannot be disabled
- Z.ai Code Bench is a private in-house benchmark
- Benchmarks vendor-reported; not independently verified
GLM-5.3 (2026-08-18) is Zhipu’s flagship text model: a 744B-total / 40B-active open-weight model with a 1M-token context and 128K max output. It shares its base model with GLM-5.2; all gains come from post-training, with a claimed 50% coding improvement on Z.ai Code Bench. Reasoning is always enabled (low/high/max, default max).
International pricing is $1.40 input / $4.40 output per 1M tokens with cached input at $0.26 (as of 2026-09-20); the China platform lists ¥8 / ¥28 with cached hits at ¥2. Open weights (FP8 and BF16) are on Hugging Face and ModelScope under Apache-2.0 (per GitHub repo metadata).
Provider
Pricing
Benchmarks with results for this model
Agents built on this model
Related models (same family)
API
Model family timeline
| Model | Released | Status |
|---|---|---|
| GLM-5.2 | 2026-06-16 | deprecated |
| GLM-5.3 (this page) | 2026-08-18 | active |
Data interpretation
| Field | Value | Evidence type |
|---|---|---|
| Context window | 1,048,576 tokens | Official |
| Architecture | 744B total / 40B active (open-weight FP8); same base model as GLM-5.2 with post-training gains | Official |
| Open weights | Yes | Official |
| License | Apache-2.0 (per GitHub repo metadata; README has no separate weights-license section - verify per-model HF cards before reuse) | Official |
| API pricing | $1.4 / $4.4 per 1M tokens (USD) | Official |
| Terminal-Bench 3.0 | 28.3 accuracy | Vendor-reported |
| DeepSWE v1.1 | 66.9 accuracy | Vendor-reported |
| Agents' Last Exam (CLI) | 28.5 accuracy | Vendor-reported |
| CyberGym | 84.5 accuracy | Vendor-reported |
| Family position | 2 models in the GLM-5 family | China AI Hub analysis |
Evidence types: Official = vendor documentation, pricing pages or model cards. Vendor-reported = benchmark scores published by the vendor. China AI Hub analysis = derived from the database itself. See the sourcing policy.
Sources
Confidence and source hierarchy per the sourcing policy. Facts change; check the source before relying on this page.
What is the context window of GLM-5.3?
GLM-5.3 has a 1,048,576-token context window and a maximum output of 131,072 tokens.
Is GLM-5.3 open weight?
Yes — GLM-5.3 weights are openly available under the Apache-2.0 (per GitHub repo metadata; README has no separate weights-license section - verify per-model HF cards before reuse) license.
How much does GLM-5.3 cost through the API?
$1.4 per 1M input tokens and $4.4 per 1M output tokens (USD).
Where does China AI Hub get its GLM-5.3 data?
From 3 sources (official pages first), last verified 2026-09-20. Benchmark scores are labeled by source type; see the sourcing policy for details.