Z.ai Unleashes GLM-4.7, Redefining Enterprise AI Development
December 31, 2025, 3:43 am
Z.ai announced GLM-4.7, its new flagship large language model, on December 22, 2025. This iteration targets real-world development workflows. It offers superior performance in complex coding, multi-step tasks, and agentic execution. GLM-4.7 achieves leading benchmarks among open-source models for tool use and consistency. It rivals proprietary leaders like Claude Sonnet 4.5. This launch reinforces Z.ai's status as "China's OpenAI." The company also aims for a historic IPO, poised to become the world's first publicly listed large-model entity. Z.ai demonstrates remarkable revenue growth, doubling annually for three years. GLM-4.7 is now the default for the GLM Coding Plan. Its open-source nature promotes wide ecosystem integration, supporting advanced AI applications and developer productivity. This model marks a significant leap in AI capabilities for enterprise and production environments.
Z.ai pushes the boundaries of artificial intelligence. The company unveiled GLM-4.7. This new large language model (LLM) targets real-world development environments. It promises stability and consistency for complex tasks. Developers demand robust AI solutions. GLM-4.7 answers that call.
This iteration significantly advances the GLM large language model family. It builds on GLM-4.6. Key improvements focus on production workflows. The model handles multi-step tasks efficiently. It supports frequent tool use. Higher demands for reliability are met.
GLM-4.7 excels in coding. Its robust support for coding workflows stands out. Complex reasoning capabilities are enhanced. Agentic-style execution shows greater consistency. The model performs stably with external tools. This makes GLM-4.7 a reliable daily production tool.
Performance metrics confirm its strength. On BrowseComp, GLM-4.7 scores 67.5. This benchmark assesses web-based tasks. For interactive tool use, it scores 87.4 on τ²-Bench. This marks the highest result among publicly available open-source models. The model ranks #1 among open models on WebDev. It holds the #6 spot overall.
Coding benchmarks showcase its prowess. SWE-bench Verified, LiveCodeBench v6, and Terminal Bench 2.0 all see strong results. GLM-4.7 performs at or above Claude Sonnet 4.5. It demonstrates clear improvements over its predecessor, GLM-4.6. Code Arena, a large-scale blind evaluation platform, further confirms its lead. GLM-4.7 ranks first among open-source models. It leads all models developed in China.
Z.ai designed GLM-4.7 for practical engineering. It addresses long-running task execution. Stable tool calling is a priority. Multi-step reasoning capabilities are crucial. These features are vital for agent-based AI systems. Developers increasingly deploy LLMs in complex scenarios. GLM-4.7 provides essential capabilities.
The model offers more predictable reasoning. Fine-grained control over complex tasks is now possible. AI systems integrate deeply into production. Such capabilities become paramount. GLM-4.7 maintains consistency across multiple interactions. It adjusts reasoning depth based on task complexity. This makes its behavior within agentic systems more reliable.
Front-end generation also sees major gains. GLM-4.7 understands visual structure better. It follows established design conventions. Tasks like generating web pages or presentation materials improve. The model produces consistent spacing. It creates clearer hierarchy. Styling appears more coherent. Manual revisions decrease significantly.
General capabilities are broadened. Conversational quality has improved. The writing style is enhanced. This expands GLM-4.7's use cases. Creative and interactive applications now benefit. The model evolves towards a coherent, open-source AI system. This signifies a holistic development approach.
Z.ai evaluated GLM-4.7 rigorously. One hundred real programming tasks were used. These tests occurred in Claude Code-based development environments. Scenarios covered front-end, back-end, and instruction-following. The new model achieved higher task completion rates. Its behavioral consistency also increased. This reduces prompt adjustments. Developers focus more on delivery. GLM-4.7 is now the default for the GLM Coding Plan.
Ecosystem integration is robust. GLM-4.7 is accessible via the BigModel.cn API. It fully integrates into the Z.ai full-stack development environment. Partners worldwide adopt the GLM Coding Plan. Platforms like TRAE, Cerebras, YouWare, Vercel, OpenRouter, and CodeBuddy use it. Widespread adoption signals its value. It embeds deeply into engineering and product workflows.
Z.ai positions itself for global leadership. The company aims for a historic IPO. It plans to list on the Stock Exchange of Hong Kong. This makes Z.ai the "World's First Large-Model Public Company." Its core business centers on independent AGI foundation model development. This is a landmark event for capital markets.
The company shows impressive financial growth. Revenue reached 57.4 million RMB in 2022. It hit 124.5 million RMB in 2023. By 2024, revenue stood at 312.4 million RMB. This represents a 130% compound annual growth rate. Revenue for the first half of 2025 was 190 million RMB. Z.ai doubled its revenue for three consecutive years. Large-model related business drives this growth.
Z.ai was founded in 2019. It originated from Tsinghua University's technological achievements. Its team pioneered large-model research in China. The company built a full-stack model portfolio. This includes language, code, multimodality, and intelligent agents. Its models are compatible with over 40 domestic chips. Z.ai maintains top-tier global technical standards.
GLM-4.7 represents a significant leap for enterprise AI. It empowers developers with advanced tools. Its open-source nature fosters innovation. Z.ai's strategic vision is clear. It targets both technical excellence and market leadership. The future of AI development looks bright with GLM-4.7 at the forefront.
Z.ai pushes the boundaries of artificial intelligence. The company unveiled GLM-4.7. This new large language model (LLM) targets real-world development environments. It promises stability and consistency for complex tasks. Developers demand robust AI solutions. GLM-4.7 answers that call.
This iteration significantly advances the GLM large language model family. It builds on GLM-4.6. Key improvements focus on production workflows. The model handles multi-step tasks efficiently. It supports frequent tool use. Higher demands for reliability are met.
GLM-4.7 excels in coding. Its robust support for coding workflows stands out. Complex reasoning capabilities are enhanced. Agentic-style execution shows greater consistency. The model performs stably with external tools. This makes GLM-4.7 a reliable daily production tool.
Performance metrics confirm its strength. On BrowseComp, GLM-4.7 scores 67.5. This benchmark assesses web-based tasks. For interactive tool use, it scores 87.4 on τ²-Bench. This marks the highest result among publicly available open-source models. The model ranks #1 among open models on WebDev. It holds the #6 spot overall.
Coding benchmarks showcase its prowess. SWE-bench Verified, LiveCodeBench v6, and Terminal Bench 2.0 all see strong results. GLM-4.7 performs at or above Claude Sonnet 4.5. It demonstrates clear improvements over its predecessor, GLM-4.6. Code Arena, a large-scale blind evaluation platform, further confirms its lead. GLM-4.7 ranks first among open-source models. It leads all models developed in China.
Z.ai designed GLM-4.7 for practical engineering. It addresses long-running task execution. Stable tool calling is a priority. Multi-step reasoning capabilities are crucial. These features are vital for agent-based AI systems. Developers increasingly deploy LLMs in complex scenarios. GLM-4.7 provides essential capabilities.
The model offers more predictable reasoning. Fine-grained control over complex tasks is now possible. AI systems integrate deeply into production. Such capabilities become paramount. GLM-4.7 maintains consistency across multiple interactions. It adjusts reasoning depth based on task complexity. This makes its behavior within agentic systems more reliable.
Front-end generation also sees major gains. GLM-4.7 understands visual structure better. It follows established design conventions. Tasks like generating web pages or presentation materials improve. The model produces consistent spacing. It creates clearer hierarchy. Styling appears more coherent. Manual revisions decrease significantly.
General capabilities are broadened. Conversational quality has improved. The writing style is enhanced. This expands GLM-4.7's use cases. Creative and interactive applications now benefit. The model evolves towards a coherent, open-source AI system. This signifies a holistic development approach.
Z.ai evaluated GLM-4.7 rigorously. One hundred real programming tasks were used. These tests occurred in Claude Code-based development environments. Scenarios covered front-end, back-end, and instruction-following. The new model achieved higher task completion rates. Its behavioral consistency also increased. This reduces prompt adjustments. Developers focus more on delivery. GLM-4.7 is now the default for the GLM Coding Plan.
Ecosystem integration is robust. GLM-4.7 is accessible via the BigModel.cn API. It fully integrates into the Z.ai full-stack development environment. Partners worldwide adopt the GLM Coding Plan. Platforms like TRAE, Cerebras, YouWare, Vercel, OpenRouter, and CodeBuddy use it. Widespread adoption signals its value. It embeds deeply into engineering and product workflows.
Z.ai positions itself for global leadership. The company aims for a historic IPO. It plans to list on the Stock Exchange of Hong Kong. This makes Z.ai the "World's First Large-Model Public Company." Its core business centers on independent AGI foundation model development. This is a landmark event for capital markets.
The company shows impressive financial growth. Revenue reached 57.4 million RMB in 2022. It hit 124.5 million RMB in 2023. By 2024, revenue stood at 312.4 million RMB. This represents a 130% compound annual growth rate. Revenue for the first half of 2025 was 190 million RMB. Z.ai doubled its revenue for three consecutive years. Large-model related business drives this growth.
Z.ai was founded in 2019. It originated from Tsinghua University's technological achievements. Its team pioneered large-model research in China. The company built a full-stack model portfolio. This includes language, code, multimodality, and intelligent agents. Its models are compatible with over 40 domestic chips. Z.ai maintains top-tier global technical standards.
GLM-4.7 represents a significant leap for enterprise AI. It empowers developers with advanced tools. Its open-source nature fosters innovation. Z.ai's strategic vision is clear. It targets both technical excellence and market leadership. The future of AI development looks bright with GLM-4.7 at the forefront.