Z.ai publishes GLM-4.7-FP8 compressed MoE model
Z.ai released GLM-4.7-FP8, a compressed (FP8) mixture-of-experts text generation model using the transformers library, available on Hugging Face under the MIT license with evaluation results.
PUBLISHED2025-12-22
OBSERVED2026-08-11
AGE8mo
SOURCES1
- Pipeline: text-generation via transformers
- Architecture: MoE (glm4_moe)
- Precision: compressed-tensors (FP8)
- Languages: en, zh
- Includes eval-results
- Arxiv: 2508.06471
- License: MIT
- Deployment: Azure, US region, endpoints compatible
COMMUNITY
No curated reactions recorded for this event. Facts and takes are kept in separate layers — community context is added by hand, never blended into the record above.