Zhipu AI Unveils GLM-4.7: A Powerhouse for Agentic Coding and AI Development
December 23, 2025, 9:52 am
Location: Australia, New South Wales, Concord
Employees: 51-200
Founded date: 2016
Total raised: $494M
Zhipu AI has released GLM-4.7. This advanced AI model targets agentic coding and complex problem-solving. It features "interleaved thinking" for enhanced stability in long tasks. Performance metrics are strong. GLM-4.7 excels in coding benchmarks like SWE-bench and LiveCodeBench. It also leads in math (AIME 2025) and complex reasoning (HLE), rivaling top Western models. With a large 200K token context, GLM-4.7 is highly capable. Its $6 monthly price point offers unparalleled value. Developers can access it through APIs, integrations, and public open-source weights. This launch empowers coders with an affordable, high-performance AI assistant.
A new force has emerged in artificial intelligence. Zhipu AI, a leading developer, just launched its GLM-4.7 model. This release marks a significant milestone. It redefines capabilities for agentic coding and complex reasoning tasks. Developers now have a powerful new tool. GLM-4.7 positions itself as a strong contender. It challenges established Western AI models.
The core of GLM-4.7's innovation lies in its "interleaved thinking" approach. This system allows the model to reason before each action. It maintains consistent logic throughout complex tasks. Traditional models often drift or repeat steps. GLM-4.7 avoids these pitfalls. It ensures stability for long chains of commands. This is crucial for extensive terminal and IDE sessions. The model's reasoning stays intact across multiple interactions. This prevents context loss. It eliminates redundant problem-solving. Developers gain more reliable automation. Complex workflows become smoother.
Performance benchmarks illustrate GLM-4.7's dominance. The model shows substantial gains in key coding metrics. SWE-bench Verified scores jumped 5.8 points to 73.8%. SWE-bench Multilingual saw a 12.9-point increase, reaching 66.7%. Terminal Bench 2.0 improved by 16.5 points, hitting 41%. These are not minor adjustments. They represent significant leaps in coding proficiency.
Further testing solidifies its position. On LiveCodeBench V6, GLM-4.7 scored 84.8 points. This performance surpassed Claude Sonnet 4.5. This benchmark measures real-world coding ability. Zhipu AI demonstrates a model truly optimized for programming. Its coding capabilities are formidable.
GLM-4.7 also shines in mathematical reasoning. The model excelled on the AIME 2025 benchmark. It achieved the best result among open-source models. It even outperformed Claude Sonnet 4.5 and GPT-5.1 in this domain. This signals a broad analytical prowess. Complex calculations and problem-solving become more accessible. Its HLE benchmark score reached 42.8%. This is nearly 12 points higher than its predecessor. It approaches results seen from GPT-5.1. The model handles intricate logic with impressive accuracy.
Tool usage has seen significant enhancements. GLM-4.7 exhibits sharp improvements on τ²-bench and browser-based tasks. It makes fewer unnecessary calls. Its actions are more deliberate and meaningful. This translates to better practical application. Automation becomes more efficient. Developers can rely on its tool integration. It smoothly interacts with various programming environments.
Affordability sets GLM-4.7 apart. It offers a GLM Coding Plan for just $6 per month. Zhipu AI claims this is seven times cheaper than comparable Western solutions. This price point makes advanced AI accessible. Small teams and individual developers can leverage its power. High-performance AI is no longer exclusive. It becomes a budget-friendly option. This democratizes sophisticated coding assistance.
Accessibility is a cornerstone of this release. GLM-4.7 integrates with popular agentic environments. These include Claude Code, Roo Code, Cline, and OpenCode. Developers can quickly incorporate it into existing workflows. Its API is available through Z.ai and OpenRouter. This ensures broad platform compatibility. Public weights are also available on Hugging Face. This allows local deployment via vLLM or SGLang. A free trial is offered via Chat.Z.Ai. Zhipu AI ensures widespread adoption.
The model also boasts an expansive context window. It handles up to 200,000 tokens. Its maximum output extends to 128,000 tokens. This capacity supports very long and complex tasks. Developers can feed extensive codebases or documentation. The AI retains context throughout. This prevents fragmentation of thought. It ensures comprehensive analysis.
Beyond coding, GLM-4.7 demonstrates versatility. It significantly improves UI and "vibe coding" capabilities. The model generates cleaner web pages. Layouts are more accurate. Visual hierarchy is improved. This speeds up rapid prototyping. Internal tool development benefits greatly. Presentation generation also sees strides. Adaptation to 16:9 format rose from 52% to 91%. The model delivers more natural dialogues. It provides direct answers. It avoids excessive filler.
GLM-4.7 represents Zhipu AI's rapid innovation cycle. It is the fourth major update to the series this year. Previous versions, GLM-4.6 and GLM-4.5, were released in September and July. This aggressive development schedule highlights Zhipu AI's commitment. They continually push AI boundaries. Each iteration builds on prior successes.
This launch impacts the AI landscape profoundly. GLM-4.7 provides a high-performance, cost-effective solution. It empowers developers globally. It enhances productivity. It accelerates software creation. Zhipu AI has delivered a true game-changer. The future of AI-assisted development looks brighter.
A new force has emerged in artificial intelligence. Zhipu AI, a leading developer, just launched its GLM-4.7 model. This release marks a significant milestone. It redefines capabilities for agentic coding and complex reasoning tasks. Developers now have a powerful new tool. GLM-4.7 positions itself as a strong contender. It challenges established Western AI models.
The core of GLM-4.7's innovation lies in its "interleaved thinking" approach. This system allows the model to reason before each action. It maintains consistent logic throughout complex tasks. Traditional models often drift or repeat steps. GLM-4.7 avoids these pitfalls. It ensures stability for long chains of commands. This is crucial for extensive terminal and IDE sessions. The model's reasoning stays intact across multiple interactions. This prevents context loss. It eliminates redundant problem-solving. Developers gain more reliable automation. Complex workflows become smoother.
Performance benchmarks illustrate GLM-4.7's dominance. The model shows substantial gains in key coding metrics. SWE-bench Verified scores jumped 5.8 points to 73.8%. SWE-bench Multilingual saw a 12.9-point increase, reaching 66.7%. Terminal Bench 2.0 improved by 16.5 points, hitting 41%. These are not minor adjustments. They represent significant leaps in coding proficiency.
Further testing solidifies its position. On LiveCodeBench V6, GLM-4.7 scored 84.8 points. This performance surpassed Claude Sonnet 4.5. This benchmark measures real-world coding ability. Zhipu AI demonstrates a model truly optimized for programming. Its coding capabilities are formidable.
GLM-4.7 also shines in mathematical reasoning. The model excelled on the AIME 2025 benchmark. It achieved the best result among open-source models. It even outperformed Claude Sonnet 4.5 and GPT-5.1 in this domain. This signals a broad analytical prowess. Complex calculations and problem-solving become more accessible. Its HLE benchmark score reached 42.8%. This is nearly 12 points higher than its predecessor. It approaches results seen from GPT-5.1. The model handles intricate logic with impressive accuracy.
Tool usage has seen significant enhancements. GLM-4.7 exhibits sharp improvements on τ²-bench and browser-based tasks. It makes fewer unnecessary calls. Its actions are more deliberate and meaningful. This translates to better practical application. Automation becomes more efficient. Developers can rely on its tool integration. It smoothly interacts with various programming environments.
Affordability sets GLM-4.7 apart. It offers a GLM Coding Plan for just $6 per month. Zhipu AI claims this is seven times cheaper than comparable Western solutions. This price point makes advanced AI accessible. Small teams and individual developers can leverage its power. High-performance AI is no longer exclusive. It becomes a budget-friendly option. This democratizes sophisticated coding assistance.
Accessibility is a cornerstone of this release. GLM-4.7 integrates with popular agentic environments. These include Claude Code, Roo Code, Cline, and OpenCode. Developers can quickly incorporate it into existing workflows. Its API is available through Z.ai and OpenRouter. This ensures broad platform compatibility. Public weights are also available on Hugging Face. This allows local deployment via vLLM or SGLang. A free trial is offered via Chat.Z.Ai. Zhipu AI ensures widespread adoption.
The model also boasts an expansive context window. It handles up to 200,000 tokens. Its maximum output extends to 128,000 tokens. This capacity supports very long and complex tasks. Developers can feed extensive codebases or documentation. The AI retains context throughout. This prevents fragmentation of thought. It ensures comprehensive analysis.
Beyond coding, GLM-4.7 demonstrates versatility. It significantly improves UI and "vibe coding" capabilities. The model generates cleaner web pages. Layouts are more accurate. Visual hierarchy is improved. This speeds up rapid prototyping. Internal tool development benefits greatly. Presentation generation also sees strides. Adaptation to 16:9 format rose from 52% to 91%. The model delivers more natural dialogues. It provides direct answers. It avoids excessive filler.
GLM-4.7 represents Zhipu AI's rapid innovation cycle. It is the fourth major update to the series this year. Previous versions, GLM-4.6 and GLM-4.5, were released in September and July. This aggressive development schedule highlights Zhipu AI's commitment. They continually push AI boundaries. Each iteration builds on prior successes.
This launch impacts the AI landscape profoundly. GLM-4.7 provides a high-performance, cost-effective solution. It empowers developers globally. It enhances productivity. It accelerates software creation. Zhipu AI has delivered a true game-changer. The future of AI-assisted development looks brighter.