GLM-5.3
Z.aiZ.ai's flagship open-weights frontier model family, built for long-horizon agentic and coding tasks. GLM-5.3, the current release, is a post-trained update of the GLM-5.2 base: a sparse MoE of about 744B total parameters (40B active) with a 1M-token context, using the 'IndexShare' sparse attention introduced in GLM-5.2. A smaller GLM-5.3-Flash ships beside it on a newly trained base with hybrid sparse and linear attention.
GLM-5.3's weights ship under Z.ai's custom GLM-5.3 License, where GLM-5.2 and GLM-5.3-Flash use plain MIT. Consolidated on 2026-07-29 from glm-4, glm-4-6, glm-4-9b, glm-5-1 and glm-5-2; openness follows GLM-5.3, the release that currently governs. See docs/reference/identity.md. The GLM-5.3 checkpoints are not yet among the declared artifacts, so the adoption reading still sums GLM-4 through GLM-5.2.
Openness
3 high confidence- weights
- open(GLM-5.3 License)
- data
- closed
- code
- partial(inference and deployment code only)
- license
- GLM-5.3-License(commercial use permitted, except that a model-as-a-service operator above USD 10 billion in annual revenue must first pass a Z.ai security review)
- sibling-sku
- GLM-5.3-Flash(a smaller, separately trained model in the same release, MIT)
GLM-5.3, the current release, ships its weights under the custom GLM-5.3 License. It reads like MIT with one bound: a company running a model-as-a-service business with more than USD 10 billion in annual revenue must pass a Z.ai security review before any commercial use. The training data is closed and only inference and deployment code is published. GLM-5.2 and the smaller GLM-5.3-Flash ship under plain MIT, and the GLM-4 releases carried the GLM-4 License.
- https://huggingface.co/api/models/zai-org/GLM-5.2 recorded 2026-08-13
card license `mit`, `gated: false`, `private: false`, 2,662,328 downloads and 4,944 likes; no corpus declared on the card.
- https://huggingface.co/zai-org/GLM-5.2/blob/main/LICENSE recorded 2026-08-13
"MIT License / Copyright (c) 2026 Zhipu AI / Permission is hereby granted, free of charge ... without restriction" - the unmodified MIT text.
- https://huggingface.co/zai-org/GLM-5.3-Flash/raw/main/LICENSE recorded 2026-10-08
GLM-5.3-Flash ships under the plain MIT text: "MIT License" and "Copyright (c) 2026 Z.AI Co., Ltd".
- https://huggingface.co/zai-org/GLM-5.3/raw/main/LICENSE recorded 2026-10-08
The "GLM-5.3 License" grants rights "to deal in the Software without restriction" but requires that a Model as a Service operator whose revenue "exceeds 10 billion US dollars" must "pass Z.AI's security review before using the Software or its derivative works for any commercial purpose".
- https://huggingface.co/zai-org/GLM-5.3/raw/main/README.md recorded 2026-10-08
Model card front matter declares "license: other" and "license_name: glm-5.3" and no training dataset; the card says "GLM-5.3 uses the same base model as GLM-5.2" and that every gain comes from post-training.
- https://raw.githubusercontent.com/zai-org/GLM-5/main/README.md recorded 2026-10-08
The GLM-5 repository README covers inference and deployment for the family: "GLM-5.3 uses the same base model as GLM-5.2" and its download table links "https://huggingface.co/zai-org/GLM-5.3".
Adoption
4 high confidenceAdoption is measured as Hugging Face downloads summed across four declared checkpoints spanning GLM-4 through GLM-5.2. GLM-5.2, the current release, accounts for nearly all of that total; the older THUDM/glm-4-9b repository now redirects to zai-org.
- https://huggingface.co/api/models/THUDM/glm-4-9b recorded 2026-08-14
11,688 downloads in the trailing 30 days for THUDM/glm-4-9b (now served as zai-org/glm-4-9b)
- https://huggingface.co/api/models/zai-org/GLM-4.6 recorded 2026-08-14
28,016 downloads in the trailing 30 days for zai-org/GLM-4.6; 1,230 likes
- https://huggingface.co/api/models/zai-org/GLM-5.1 recorded 2026-08-14
72,174 downloads in the trailing 30 days for zai-org/GLM-5.1; 1,839 likes
- https://huggingface.co/api/models/zai-org/GLM-5.2 recorded 2026-08-14
2,662,328 downloads in the trailing 30 days for zai-org/GLM-5.2; 4,948 likes; gated false
- https://huggingface.co/zai-org/GLM-4-9B-Chat recorded 2026-06-04
GLM-4-9B-Chat model card: 121,169 downloads last month; the undeclared, generation-behind SKU the earlier reading came from
Capability
5 high confidenceGLM-5.3 is among the highest-scoring large open-weight models on Artificial Analysis's Intelligence Index, well ahead of GLM-5.2. Z.ai's own card puts it close to leading proprietary models on Terminal-Bench 2.1, though it trails them on DeepSWE.
- https://artificialanalysis.ai/models/glm-5-2 recorded 2026-09-29
Artificial Analysis page for the predecessor: "GLM-5.2 (max) scores 34 on the Artificial Analysis Intelligence Index", below GLM-5.3.
- https://artificialanalysis.ai/models/glm-5-3 recorded 2026-09-29
Artificial Analysis page for the current release: "GLM-5.3 (max) scores 45 on the Artificial Analysis Intelligence Index, placing it well above average among comparable models" and "is amongst the leading models in intelligence".
- https://huggingface.co/zai-org/GLM-5.3/raw/main/README.md recorded 2026-09-29
GLM-5.3 model card benchmark table: vendor-reported "Terminal Bench 2.1 | 88.2" and "DeepSWE (v1.1) | 66.9" for GLM-5.3, against 81.0 and 46.2 for GLM-5.2.
- https://www.vals.ai/models/zai_glm-5.1-thinking recorded 2026-08-13
The page returns HTTP 404 - it establishes nothing and the reading does not depend on it.
Verified 2026-09-29, except some scores, still awaiting re-confirmation