Zhipu · 2× · tolest 27. Aug. 2026
GLM-5.2 is the current open-weight flagship language model from Zhipu AI (brand name Z.ai), released as the successor to GLM-5.1 in the GLM-5.X model series on June 13/16, 2026. It is a Mixture-of-Experts model with approximately 744 billion total parameters and roughly 40 billion active parameters per Token, designed for long-running coding and agent workflows with a practically usable 1-million-Token context window. The weights are freely available under the MIT license on Hugging Face; a direct successor (GLM-5.3) was initially released only via API in August 2026, without open weights. According to the official technical report, the model achieves leading results among open models on Benchmarks including SWE-bench Pro and Terminal-Bench 2.1.
Features
| Key Benchmark (%) | SWE-bench Pro: 62.1 (vs. GLM-5.1: 58.4); Terminal-Bench 2.1: 81.0 |
| Context Window (Tokens) | 1,048,576 tokens (1M), max output 128K tokens |
| License | MIT license (open weights, no regional restrictions) |
| Multimodality | Text-only (text input/output); no native vision/image support |
| Platform | Hugging Face (zai-org/GLM-5.2), Z.ai API platform; deployable via SGLang, vLLM, Transformers, KTransformers, Unsloth, Ascend NPU |
| Price per 1M Tokens | $1.40 input / $4.40 output (Z.ai official API pricing); cached input $0.26 |
| Release Date | June 13, 2026 (Coding Plan preview), June 16, 2026 (open weights) |