Futures
Access hundreds of perpetual contracts
TradFi
Gold
One platform for global traditional assets
Options
Hot
Trade European-style vanilla options
Unified Account
Maximize your capital efficiency
Demo Trading
Introduction to Futures Trading
Learn the basics of futures trading
Futures Events
Join events to earn rewards
Demo Trading
Use virtual funds to practice risk-free trading
Launch
CandyDrop
Collect candies to earn airdrops
Launchpool
Quick staking, earn potential new tokens
HODLer Airdrop
Hold GT and get massive airdrops for free
Launchpad
Be early to the next big token project
Alpha Points
Trade on-chain assets and earn airdrops
Futures Points
Earn futures points and claim airdrop rewards
Times Observer | Cost and Technology Synergy: Chinese AI Breaking Through in Competition
Securities Times Reporter Chen Xiachang
MiniMax M2.5 has ranked first in global large model call volume for five consecutive weeks, and Moon’s Dark Side Kimi K2.5 has become a core support for the well-known programming platform Cursor. These recent news items confirm that domestic open-source large models have moved from technical follow-up to leading the race. In the increasingly fierce global AI competition, Chinese companies are leveraging precise cost control and continuous technological iteration to forge a uniquely innovative path.
Cost advantage is the foundation for domestic open-source large models to break through and the core confidence in winning the global market. The cost advantage of domestic large models stems from their unique energy supply advantages. Data shows that electricity costs account for 70% to 80% of computing power expenses, and even slight differences in electricity prices can translate into huge operational cost disparities. Relying on a stable energy supply system and relatively low electricity prices, Chinese AI companies have built a solid cost defense line.
Technological innovation fundamentally amplifies the cost advantage and has become the core engine driving the leadership of domestic open-source large models. Unlike some overseas models that pursue parameter scale, domestic open-source large models adhere to an “efficient and practical” technical route, achieving performance breakthroughs through underlying architecture innovations. MiniMax M2.5 optimizes architecture to complete the same tasks with fewer tokens, reducing inference costs from the source and balancing performance and efficiency. Moon’s Dark Side, through deep technical cultivation, has taken a differentiated path—enhancing core capabilities such as coding and visual understanding, and achieving multiple-fold efficiency improvements.
The collaborative development of the open-source ecosystem further consolidates the dual advantages of cost and technology. Domestic open-source large model companies abandon the “closed-door” approach and actively promote ecosystem co-construction. The licensing cooperation between Kimi K2.5 and Cursor exemplifies this open-source ecosystem collaboration. Such cooperation not only enables efficient technological implementation but also, through resource sharing and complementary strengths, further spreads R&D and operational costs, accelerating technological iteration.
The current position of domestic open-source large models is not achieved through a single-dimensional breakthrough but through systematic success in cost control, technological innovation, and ecosystem co-building. Of course, the path of innovation still faces challenges, such as deep originality of core technologies and the规范 development of the open-source ecosystem—all areas requiring ongoing efforts. However, it is undeniable that the development path of cost foundation and technological empowerment has enabled domestic open-source large models to take the lead in global competition.
(Edited by: Zhang Xiaobo)
【Disclaimer】This article only reflects the author’s personal views and is not related to Hexun.com. Hexun.com maintains neutrality regarding the statements and opinions in this article and does not provide any explicit or implicit guarantees regarding the accuracy, reliability, or completeness of the content. Readers are advised to use it as a reference and bear all responsibilities themselves. Email: news_center@staff.hexun.com