GLM-5.3
Canonical page on the main site: chinaaihub.com/models/glm-53
Provider
Model Family
GLM-5
Release Date
2026-08-18
Status
active
Architecture
744B total / 40B active (open-weight FP8); same base model as GLM-5.2 with post-training gains
Parameter Information
- total_parameters: 744B
- active_parameters: 40B
- parameter_precision: FP8 (BF16 variant also released)
Context Window
1048576
Maximum Output
131072
Capabilities
- reasoning: Yes
- coding: Yes
- vision: No
Open Weight
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
Pricing
- input_price_per_1m: 1.4
- output_price_per_1m: 4.4
- currency: USD
- pricing_ref: zhipu-ai
Official Api
Yes
Cloud Providers
- Z.ai
- BigModel
Regions
- international
- china
Benchmark Results
| benchmark | score | metric | date | source_type | source_url |
|---|---|---|---|---|---|
| Terminal-Bench 3.0 | 28.3 | accuracy | 2026-08-18 | vendor_reported | https://docs.z.ai/guides/llm/glm-5.3 |
| DeepSWE v1.1 | 66.9 | accuracy | 2026-08-18 | vendor_reported | https://docs.z.ai/guides/llm/glm-5.3 |
| Agents' Last Exam (CLI) | 28.5 | accuracy | 2026-08-18 | vendor_reported | https://docs.z.ai/guides/llm/glm-5.3 |
| CyberGym | 84.5 | accuracy | 2026-08-18 | vendor_reported | https://docs.z.ai/guides/llm/glm-5.3 |
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
Last Verified
2026-09-20
Sources
| source_name | source_url | source_type | last_verified | confidence |
|---|---|---|---|---|
| Z.ai docs — GLM-5.3 model page | https://docs.z.ai/guides/llm/glm-5.3 | official | 2026-09-20 | high |
| Z.ai pricing | https://docs.z.ai/guides/overview/pricing | official | 2026-09-20 | high |
| GLM-5 GitHub repository | https://github.com/zai-org/GLM-5 | official | 2026-09-20 | high |