Zhipu launches open-source GLM-5.2 with 1M context window
- Source
- Z.ai
- Time
- 6:50 PM
- Weight
- 94/100
Zhipu has announced the release of GLM-5.2, its latest flagship open-source model designed specifically for long-horizon engineering and coding tasks. The model features a 1M-token context window that is engineered for stability and reliability during complex, multi-step trajectories like large-scale debugging and automated research.
Available under an MIT license, GLM-5.2 introduces a new architecture called IndexShare, which reduces computational overhead by reusing indexers across attention layers, improving performance while cutting per-token FLOPs by nearly threefold at its maximum context length. In performance evaluations, GLM-5.2 has emerged as a top-tier open-source model, rivaling closed-source frontiers such as GPT-5.5 and Claude Opus 4.8 on benchmarks like FrontierSWE and SWE-bench Pro.