Z.ai Launches GLM-5.1 Autonomous Coding AI Model
Z.ai has released GLM-5.1, an open-source flagship model designed for agentic engineering and long-duration autonomous coding. With this launch, the company advances capabilities that allow a single model to manage complex coding workflows for up to eight hours continuously.
At the same time, GLM-5.1 achieves a leading score of 58.4 on the SWE-Bench Pro benchmark. As a result, it surpasses competing models from OpenAI, Anthropic, and Google on that test.
Extended autonomous coding and performance gains
GLM-5.1 builds on the earlier GLM-5 architecture introduced in February. However, it enhances coding performance through progressive alignment techniques, including multi-task supervised fine-tuning and reinforcement learning stages. Therefore, the model executes long, multi-step tasks with improved consistency and precision.
In addition, the system supports continuous execution loops that combine planning, coding, testing, and optimization. Consequently, it can complete complex development workflows without interruption. For example, the model can run hundreds of iterative cycles while refining outputs in real time.
Moreover, GLM-5.1 offers a 200,000-token context window with up to 128,000 output tokens. As a result, it handles large-scale coding environments and extended problem-solving scenarios more effectively. On KernelBench Level 3, it delivers a 3.6x speed improvement on machine learning workloads, which further strengthens its performance profile.
Real-world capabilities and workflow integration
Alongside benchmark performance, GLM-5.1 demonstrates practical autonomous execution in real-world scenarios. For instance, the model can complete full development cycles, including building systems from scratch within extended timeframes.
Furthermore, it integrates with agentic coding tools such as Claude Code and OpenClaw. Therefore, developers can incorporate the model into existing workflows without significant friction. In parallel, its optimization capabilities improve system performance across iterative development stages.
As a result, GLM-5.1 supports advanced engineering tasks that require sustained execution and continuous refinement. This positions the model as a strong option for complex coding environments and research applications.
Open-source strategy and competitive positioning
GLM-5.1 is available under an MIT license, which allows broad access and customization. At the same time, Z.ai offers API pricing at $1.00 per million input tokens and $3.20 per million output tokens. Consequently, the model remains competitive in both open-source and commercial deployment scenarios.
Meanwhile, the release intensifies competition in the open-source coding model space. GLM-5.1 now leads on SWE-Bench Pro, although performance gaps remain in broader reasoning and creative tasks when compared to top-tier models.
Overall, the launch reflects a growing shift toward open, agentic AI systems capable of sustained autonomous work. As development continues, GLM-5.1 strengthens Z.ai’s position within the evolving global AI ecosystem.
