Home
MiniMax 2.5 Tops Global Usage With Over $150M Monthly Revenue; M3 Flagship Version Set for H1 Release

Chinese large language models are generating intense momentum across the global developer ecosystem, driven by their reputation for cost-efficiency and engineering excellence.
As of March 3, 2026, data from OpenRouter, the world’s leading API aggregation platform, reveals unprecedented dominance by domestic models: MiniMax M2.5 tops global call volumes, followed closely by Kimi K2.5 and GLM-5. This “top three” lineup signals that Chinese providers have overtaken international rivals through superior efficiency in AI deployment and strong developer adoption.
3.07 Trillion Tokens in Seven Days: MiniMax’s Rapid Expansion
Released on February 13, MiniMax M2.5 processed 3.07 trillion tokens within its first week. Designed as “the world’s first native agent model,” it directly addresses key developer needs:
Agent-First Architecture: Beyond conversational interfaces, it functions as a backend engine capable of executing complex coding tasks and automated workflows.
Unbeatable Pricing: Extremely low inference costs enable developers to build high-frequency AI applications without financial constraints.
Commercial Breakthrough: Driven by M2.5’s adoption, MiniMax’s Annual Recurring Revenue (ARR) reached $150 million in February, accelerating its commercialization timeline.
Strategic Trade-Off: The Case for 10B Activated Parameters
While competitors often scale toward billions or trillions of parameters, MiniMax M2.5 uses only 10 billion activated parameters. This design prioritizes reasoning speed and practical output. By optimizing its Mixture of Experts (MoE) architecture, MiniMax demonstrates that compact models can match top-tier performance in coding logic. However, users note that its knowledge breadth remains a limitation.
Upcoming Showdown: M3.0 vs. DeepSeek V4
China’s large model market is now defined by rapid, weekly iterations:
DeepSeek’s Upcoming Release: DeepSeek V4 is scheduled for launch next week, with global attention focused on its latest cost-reduction and performance enhancements.
MiniMax M3 Series Preview: XiYu Technology plans to introduce the MiniMax M3 Series later this year.
Morgan Stanley’s Positive Outlook: According to a recent Morgan Stanley report, M3.0 will incorporate advanced pre-training and architectural improvements to address knowledge gaps, while supporting robust multimodal capabilities. Its overall performance is projected to compete directly with leading global flagship models.
Related article
Meta's Muse outpaces ChatGPT in early mobile launch
New data suggests Meta’s latest AI app, Muse, could become the social giant’s next major success. According to recent estimates from Appfigures, Muse has been downloaded on mobile devices more times in its first twelve days than ChatGPT did in the tw
TechCrunch Disrupt 2026: Clay’s Kareem Amin on the rise of the GTM engineer
Two years ago, GTM engineers were virtually nonexistent in most organizations. Today, the role is emerging across the tech sector, with professionals leveraging AI and automation to construct revenue engines that previously demanded multiple hires, d
SAS Launches Enterprise AI Governance Tool to Manage Intelligent Entities and Mitigate Shadow AI Risks
During the recent SAS Innovate conference, analytics leader SAS unveiled a suite of governance tools designed for Agentic AI. These solutions empower organizations to create transparent, secure automated ecosystems, tackling the compliance and trust
Related Special Topic Recommendations
Comments (0)
0/500

Chinese large language models are generating intense momentum across the global developer ecosystem, driven by their reputation for cost-efficiency and engineering excellence.
As of March 3, 2026, data from OpenRouter, the world’s leading API aggregation platform, reveals unprecedented dominance by domestic models: MiniMax M2.5 tops global call volumes, followed closely by Kimi K2.5 and GLM-5. This “top three” lineup signals that Chinese providers have overtaken international rivals through superior efficiency in AI deployment and strong developer adoption.
3.07 Trillion Tokens in Seven Days: MiniMax’s Rapid Expansion
Released on February 13, MiniMax M2.5 processed 3.07 trillion tokens within its first week. Designed as “the world’s first native agent model,” it directly addresses key developer needs:
Agent-First Architecture: Beyond conversational interfaces, it functions as a backend engine capable of executing complex coding tasks and automated workflows.
Unbeatable Pricing: Extremely low inference costs enable developers to build high-frequency AI applications without financial constraints.
Commercial Breakthrough: Driven by M2.5’s adoption, MiniMax’s Annual Recurring Revenue (ARR) reached $150 million in February, accelerating its commercialization timeline.
Strategic Trade-Off: The Case for 10B Activated Parameters
While competitors often scale toward billions or trillions of parameters, MiniMax M2.5 uses only 10 billion activated parameters. This design prioritizes reasoning speed and practical output. By optimizing its Mixture of Experts (MoE) architecture, MiniMax demonstrates that compact models can match top-tier performance in coding logic. However, users note that its knowledge breadth remains a limitation.
Upcoming Showdown: M3.0 vs. DeepSeek V4
China’s large model market is now defined by rapid, weekly iterations:
DeepSeek’s Upcoming Release: DeepSeek V4 is scheduled for launch next week, with global attention focused on its latest cost-reduction and performance enhancements.
MiniMax M3 Series Preview: XiYu Technology plans to introduce the MiniMax M3 Series later this year.
Morgan Stanley’s Positive Outlook: According to a recent Morgan Stanley report, M3.0 will incorporate advanced pre-training and architectural improvements to address knowledge gaps, while supporting robust multimodal capabilities. Its overall performance is projected to compete directly with leading global flagship models.
Meta's Muse outpaces ChatGPT in early mobile launch
New data suggests Meta’s latest AI app, Muse, could become the social giant’s next major success. According to recent estimates from Appfigures, Muse has been downloaded on mobile devices more times in its first twelve days than ChatGPT did in the tw
TechCrunch Disrupt 2026: Clay’s Kareem Amin on the rise of the GTM engineer
Two years ago, GTM engineers were virtually nonexistent in most organizations. Today, the role is emerging across the tech sector, with professionals leveraging AI and automation to construct revenue engines that previously demanded multiple hires, d











