GLM-5
Editorial notes
Lanzado 11 feb 2026 (API; open-weight Q1 2026). MoE 744B total / 40B activos (vs GLM-4.5: 355B/32B). Pre-training 28.5T tokens (vs 23T). Integra DeepSeek Sparse Attention (DSA). LiveBench Agentic Coding 55.0 (#1 open source). HMMT Nov 2025 96.9. IMOAnswerBench 82.5. MIT. Pricing $1.00/$3.20 per MTok. Scores del model card oficial zai-org/GLM-5 (verif. NVIDIA NIM + Ollama). Auditoria 2026-06-08: HLE corregido 50.4->30.5 (el 50.4 era HLE CON tools; sin tools = 30.5). AIME 92.7 REMOVIDO de AIME-2025: el valor real es AIME 2026 I (92.7), GLM-5 no reporta AIME 2025.
Spec sheet
- Company
- Zhipu AI
- Country
- CN
- Type
- hybrid
- Release
- 2026-02
- Context
- 200K tokens
- Slug
- glm-5
Benchmarks (3)
Cite this model
BibTeX · APA
BibTeX
@misc{frontier-glm-5,
title = {GLM-5},
author = {{Zhipu AI}},
year = {2026},
note = {Frontier Benchmarks AI atlas. Accessed 2026-06-17},
url = {https://frontierbenchmarks.com/models/glm-5}
} APA
Zhipu AI (2026). GLM-5 [Large language model]. Frontier Benchmarks AI. Retrieved 2026-06-17, from https://frontierbenchmarks.com/models/glm-5
Citation reflects the atlas page, not the original model paper. For the paper, see the "Resources" section above.