Chinese AI startup Z.ai, renowned for its open-source GLM models, has launched GLM-5.1 under the MIT License, allowing commercial use and customization. This latest model features 754 billion parameters and can autonomously work on tasks for up to eight hours, significantly exceeding prior limits. GLM-5.1 adopts a unique staircase optimization pattern, autonomously refining complex engineering problems with marked breakthroughs, such as optimizing a vector database to achieve 21,500 queries per second, a sixfold improvement over prior models. The model excels in KernelBench, demonstrating sustained gains and competitive performance. Offered through tiered subscriptions and API access, GLM-5.1 balances open source accessibility with proprietary high-speed variants like GLM-5 Turbo. Benchmarks highlight its superiority, outperforming GPT-5.4, Claude Opus 4.6, and others in coding tasks. Users commend its reliability and efficiency, reporting dramatic time savings in complex workflows. This release marks a shift toward AI systems capable of extended, autonomous work, promising to reshape software development and agentic engineering globally.
Back
