GLM-5.1 is live – coding ability on par with Claude Opus 4.5

Zhipu AI has officially released GLM-5.1, its latest flagship large language model, and it's making waves by challenging the coding supremacy of closed-source giants. The model, now available to all Coding Plan users, boasts a massive 744 billion parameters (with 40 billion activated per query) and was trained on 28.5 trillion tokens. Its standout feature is a 200,000-token context window with a 128,000-token output limit, enabling it to process and reason over entire codebases. Crucially, it includes native support for the Model Context Protocol (MCP), a standard for connecting AI models to tools and data sources, which is essential for building capable AI agents.

Benchmark results from March 2026 solidify its position: a score of 77.8 on SWE-bench-Verified, which tests models on real-world software engineering issues from open-source projects, marks it as the highest-scoring open-source model available. It also achieved 56.2 on Terminal Bench 2.0, another state-of-the-art result. These scores indicate it not only beats OpenAI's GPT-4o on coding tasks but is now competitive with Anthropic's top-tier Claude Opus 4.5. For developers, this translates to a powerful tool for autonomous, multi-step coding tasks, complex refactoring and debugging within long contexts, and orchestrating complete agentic workflows from planning to execution and delivery.

The model is accessible through Zhipu AI's platform via three Coding Plan tiers: Lite, Pro, and Max. Its arrival signals a significant shift, providing the developer community with an open-source alternative that delivers near top-tier, proprietary performance. This increases competitive pressure in the AI coding assistant space and gives teams more flexibility and control over their AI development tools without sacrificing capability.