Exclusive sources reveal that on July 11, 2026, Tang Jie, the founder of Zhipu AI, released an internal letter titled "The Great Wave has Come." Since its H-share IPO on January 8, 2026, #Zhipu's market value has surged by 10x, entering the "Trillion HKD Club" in June 2026—a figure nearly triple Baidu's market cap and exceeding Xiaomi's.
Zhipu's valuation surge stems from its decisive bet on Coding in early 2025. Tang Jie emphasized that the emergence of DeepSeek R1 marked the end of the traditional Chat paradigm exploration. The post-#DeepSeek era is defined by Coding and Reasoning—capabilities crucial for co-existing with Agents. This echoes the trajectory of Anthropic, whose ARR skyrocketed from $87 million in January 2024 to $47 billion in June 2026 due to its heavy focus on coding.
Technically, Zhipu's journey from GLM-4.5 to the open-sourced GLM-5.2 has placed it in the top tier of global AI coding models, with key benchmarks matching or surpassing Claude Opus 4.8 and GPT-5.5. This technical prowess translates to financial success: as of March 2026, Zhipu's MaaS platform ARR reached 1.7 billion RMB, a 60-fold increase year-over-year.
In his letter, Tang Jie outlined three post-coding pillars to achieve #AGI: 1) Long Horizon Tasks, enabling models to plan and execute tasks over weeks or months; 2) Autonomous Agent Systems, evolving from "One-Person Companies (OPC)" to automated "NPC companies" by solving memory, continuous learning, and self-judgment; and 3) Self-Evolving models that write code, clean data, and train themselves utilizing massive compute clusters.
[AgentUpdate Depth Analysis] Zhipu AI's strategic pivot from conversational AI to autonomous agentic systems highlights a profound shift in the AGI landscape. As the "Chat" paradigm plateaus post-DeepSeek R1, the industry is racing toward "Action." By prioritizing Long Horizon Tasks and Self-Evolving capabilities, Zhipu addresses the core bottlenecks of current AI agents: long-term execution drift and high human-in-the-loop dependencies. Comparing this with OpenAI's agentic roadmap and Anthropic's computer-use paradigms, Zhipu’s vision of transitioning from "One-Person Companies (OPC)" to "Non-Player Characters (NPC)" companies illustrates a future where agent swarms autonomously manage enterprise workflows. For the global AI Agent ecosystem, this means the value bottleneck is shifting from prompt-based orchestration layers to the foundation model's native capability to self-judge and self-evolve, significantly accelerating agent-native monetization.
