GLM 5
#52zhipu-ai · v5 · seit Februar 2026 (laut mehreren Quellen konkret 11. Februar 2026) · 32× · zuletzt 02. Juli 2026
26
Momentum
GLM-5 is the fifth-generation large language model released by Zhipu AI (Z.ai) in February 2026, designed for complex systems engineering and long-horizon agentic tasks. It uses a Mixture-of-Experts architecture with approximately 744 billion total parameters and 40 billion active parameters per token, and integrates DeepSeek Sparse Attention to reduce deployment costs at long context lengths. On SWE-bench Verified, GLM-5 officially scores 77.8%, which Z.ai documents as the leading result among open-weight models. The model weights are released under the MIT license and are accessible via the Z.ai API platform and Hugging Face.
Historique du momentum
04.04.03.07.
Fonctionnalités
| Key Benchmark (%) | SWE-bench Verified: 77.8% |
| Context Window (Tokens) | 200K (up to 203K depending on provider) |
| License | MIT License (open weights) |
| Multimodality | Base GLM-5 primarily text-based; multimodal vision/video capabilities in separate model GLM-5V-Turbo (CogViT encoder) |
| Platform | Z.ai API platform, Hugging Face, ModelScope, local deployment (vLLM/SGLang) |
| Price per 1M Tokens | from $0.573 input / $1.92 output (provider average) |
| Release Date | February 2026 (approx. Feb 11, 2026) |
