Skip to content
NEWSee what changed →

zai-org · glm4_moe

GLM-4.5-Air

GLM-4.5-Air by zai-org: a 110.5B-parameter mixture-of-experts open-weight model under the mit license.

mitMixture of experts131K contextAuto-imported · not yet reviewed

1,101,409 downloads · 650 likes · Model card · synced 2026-11-19

Specifications

Parameters
110.5B (6.9B active)
Architecture
moe · glm4_moe
Layers
46
Hidden size
4,096
KV heads · head dim
8 · 128
Vocabulary
151,552
Native dtype
bfloat16
Context window
131,072 tokens
Released
2025-07-20
Last modified on Hub
2025-08-11

Features & licensing

License
mit
Commercial use
Yes
Modalities
text
Tool / function calling
No
Reasoning mode
No
Languages
EN, ZH
Quantised variants on Hub
AWQ, FP8, GGUF, GPTQ, bitsandbytes
Pipeline
text-generation
transformerssafetensorsglm4_moetext-generationconversationalenzheval-resultsendpoints_compatible

Task fit

Editorial scores (0–100) used by the recommendation engine.

  • Chat70
  • RAG / Q&A70
  • Code70
  • Summarisation70
  • Extraction70
  • Agents70

Hardware to run GLM-4.5-Air

Weights need about 243 GB at FP16, 128 GB at INT8 and 73 GB at INT4. The KV cache adds roughly 188 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 109.9 GB, which fits on 2 × NVIDIA A100 (80 GB).

GPU configurations at the reference workload
ConfigurationVRAMUtilisationEst. first tokenCloud / month
5 × NVIDIA L4 (24 GB)120 GB92 %~0.5 s$2,920
3 × NVIDIA L40S (48 GB)144 GB76 %~0.2 s$4,161
2 × NVIDIA A100 (80 GB)recommended160 GB69 %~0.1 s$4,672
2 × NVIDIA H100 (80 GB)160 GB69 %~0.0 s$6,570

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.