On April 10, Zhipu officially released and open-sourced its new flagship model GLM-5.1. The most significant breakthrough of this model lies in its long-horizon task processing capabilities, making it the world's first open-source model capable of working continuously for 8 hours.

image.png

Core Breakthrough: From "Q&A" to "Self-Working"

Zhipu Team pointed out that the next standard for large models is no longer simply about benchmark scores, but rather "how long they can work independently."

  • 8-Hour Endurance: GLM-5.1 can work independently and continuously for more than 8 hours on a single task. During this time, it can autonomously plan, execute, identify bottlenecks, and self-evolve, ultimately delivering engineering-level results.

  • Code Capabilities Top-Notch: In the SWE-bench Pro benchmark test, which measures professional software development, GLM-5.1 achieved the first surpassing of the overseas top model Opus4.6 by a domestic model.

  • Remarkable Real-World Performance: In programming tests targeting massive data retrieval systems, GLM-5.1 underwent over 6,000 optimization operations, finally achieving a speed six times faster than the previous best performance.

image.png

Industry Trends: Moving Away from "Price Wars" to "Performance Premium"

With a leap in performance, Zhipu's pricing strategy has also undergone a major shift:

  • Price Catching Up with Overseas: Platform data shows that GLM-5.1 raised its price by 10% again. Its pricing in coding scenarios has for the first time caught up with the leading overseas manufacturer Anthropic.

  • Returning to Commercial Value: CEO Zhang Peng of Zhipu stated that long-term low-price competition is not beneficial for the industry, and the price adjustment aims to bring AI value back to a normal range. Currently, Zhipu has achieved global value monetization, with its API business ARR (Annual Recurring Revenue) increasing by 60 times year-on-year.

Market Trends: Underlying Computing Power and Models Enter a "Price Hike" Trend

Since the beginning of 2026, the domestic AI industry is undergoing a collective transformation from "low price for volume" to "value-based pricing":

  • Tencent Cloud: Announced a 5% increase in AI computing power and container services.

  • Aliyun: AI computing-related product prices increased by 5%-34%.

  • Baidu Intelligent Cloud: Related AI computing services increased by 5%-30%.

Conclusion: The Timeline for AGI

Industry consensus holds that an important indicator for measuring AGI is the timeline for task completion. The time required for cutting-edge models to complete tasks doubles every seven months. With GLM-5.1 launching its "8-hour work mode," large models are formally transitioning from chatbots that respond to questions to "virtual employees" capable of deeply participating in complex projects.