Domestic AI sparks a buying frenzy among foreigners, KNOWLEDGE ATLAS open-source model rivals Anthropic, costing only 23 yuan per month

According to reports from domestic media, China's AI large models have recently sparked a buying frenzy overseas. KNOWLEDGE ATLAS announced the open-sourcing of GLM-5.1 earlier, with programming performance comparable to leading American AIs, attracting a large number of overseas developers to purchase across the ocean due to significant regional price differences. Data shows that the weekly invocation volume of domestic AI reached as high as 12.96 trillion tokens, surpassing the United States for five consecutive weeks. Pricing Differences Between Domestic and Overseas KNOWLEDGE ATLAS officially announced the open-sourcing of GLM-5.1 on April 8. According to the model aggregation platform OpenRouter, KNOWLEDGE ATLAS has raised the price of GLM by another 10%. After the price adjustment, the token price for cache hits in programming scenarios for GLM-5.1 is now close to the level of Claude Sonnet 4.6 under Anthropic, marking the first time a domestic large model has aligned its pricing in core scenarios with leading overseas companies. Although the pricing for the overseas version has been adjusted, with monthly fees ranging from $18 to $160, the domestic version charges only about 49 to 469 RMB. The significant price difference has attracted many overseas netizens to post on social media, researching how to bind payment methods such as WeChat and Alipay, and even asking whether an ID card is needed to purchase large models like KNOWLEDGE ATLAS and MiniMax through domestic channels. The domestic second-hand platform Xianyu has also seen the emergence of the first wave of "scalpers," purchasing AI programming packages for overseas users to profit from the price difference, claiming a 100% success rate, with some sales exceeding a thousand. Analysis indicates that regional price differences are a common strategy in the global SaaS industry. This is mainly due to varying payment capabilities in different regions, as well as higher computing power and payment channel costs associated with the overseas version. The European and American markets primarily target high average revenue per user (ARPU), while the domestic market focuses on a massive engineer market, prioritizing scale before monetization. Performance Described as Crazy by Overseas Developers In addition to the price advantage, the performance of domestic large models has also been described as crazy by overseas developers. One overseas netizen stated that GLM-5.1 only costs $3 per month (approximately HKD 23.4), achieving about 94.6% of the programming performance of Claude Opus 4.6, which is simply insane. Other netizens mentioned that when tasked with complex assignments like building a simulated trading system, the model could almost complete them without any additional guidance. Additionally, some users noted that for local deployment scenarios, the biggest advantage of GLM-5.1 is its ability to resist context drift, maintaining reasoning quality even after 8 hours of continuous operation and actively adjusting strategies, making it highly practical for local automated development intelligent scenarios. Weekly Invocation Volume Exceeds 12 Trillion, Surpassing the U.S. for Five Consecutive Weeks According to the latest data from OpenRouter, during the week from March 30 to April 5, the weekly invocation volume of China's AI large models reached approximately 12.96 trillion tokens, with a weekly growth of over 31%, surpassing the United States for the fifth consecutive week. During the same period, the weekly invocation volume of U.S. AI large models was about 3.03 trillion tokens, with a slight weekly increase of about 0.76%. The top six global invocation volumes are all from Chinese AI large models, with Alibaba's (9988) Qianwen 3.6 series ranking first Data shows that in March this year, China's daily average Token call volume has exceeded 140 trillion, an increase of more than a thousand times compared to 100 billion at the beginning of 2024. Whether it is the enthusiasm in overseas markets or the performance in data, it reflects that domestic AI is moving from being a "follower" to aligning with international first-class standards, with the gap in price and performance between the two sides continuously narrowing