zai-org · glm_moe_dsa
GLM-5.2-FP8
GLM-5.2-FP8 by zai-org: a 753.3B-parameter mixture-of-experts open-weight model under the mit license.
986,105 downloads · 266 likes · Model card · synced 2026-09-19
Specifications
- Parameters
- 753.3B (23.5B active)
- Architecture
- moe · glm_moe_dsa
- Layers
- 78
- Hidden size
- 6,144
- KV heads · head dim
- 64 · 192
- Vocabulary
- 154,880
- Native dtype
- —
- Context window
- 1,048,576 tokens
- Released
- 2026-06-16
- Last modified on Hub
- 2026-09-01
Features & licensing
- License
- mit
- Commercial use
- Yes
- Modalities
- text
- Tool / function calling
- No
- Reasoning mode
- No
- Languages
- EN, ZH
- Quantised variants on Hub
- —
- Pipeline
- text-generation
Task fit
Editorial scores (0–100) used by the recommendation engine.
- Chat70
- RAG / Q&A70
- Code70
- Summarisation70
- Extraction70
- Agents70
Hardware to run GLM-5.2-FP8
Weights need about 1657 GB at FP16, 870 GB at INT8 and 497 GB at INT4. The KV cache adds roughly 3834 MB per 1,000 tokens per request. For a reference workload of 5 requests per second with 1,500 input and 300 output tokens, total GPU memory is around 2153.5 GB.
| Configuration | VRAM | Utilisation | Est. first token | Cloud / month |
|---|
Estimates only. Run an assessment for your own traffic.
Auto-imported from Hugging Face. Architecture, license and feature data are synced from the model repository; the task-fit scores below are provisional defaults until a curator reviews them.