Home
MiniMax M2.5 prices at one-tenth of overseas models, tops global API call volume for five straight weeks

Chinese large models are delivering a "dimensional strike" in the global computing power market through extreme cost-effectiveness.
According to the latest report, the Chinese large model MiniMax M2.5 has topped global large model usage rankings for five consecutive weeks. This achievement not only sets a new record for Chinese model activity on the international stage but also highlights China's core competitive advantage in the global AI industry: domestic models are priced dozens of times lower than overseas competitors with comparable capabilities.
This cost advantage, which leaves competitors struggling, stems from two strategic moats:
Underlying architectural innovation: making tokens more "valuable"
MiniMax developers note that fundamental innovations in the underlying architecture have significantly reduced inference costs. In simple terms, improved algorithm efficiency allows the model to complete the same complex tasks using fewer tokens, directly lowering users' computing costs at the technical level.
Energy and infrastructure advantages: building a computing power defense line
Beyond technological optimizations, China's energy advantage in computing power infrastructure cannot be overlooked. Industry data shows that electricity costs account for 70% to 80% of computing costs. For AI clusters running millions of graphics cards in parallel, China's stable energy supply and relatively low industrial electricity prices provide a highly competitive "cost firewall" for the AI industry.
From "computing power input" to "intellectual output," Chinese large models represented by MiniMax are rapidly capturing global market share through the dual forces of technological iteration and cost control. As AI competition enters its second half, this "high intelligence, low price" model may completely rewrite the rules of global large model commercialization.
Related article
OpenAI Unveils Codex Referral Program to Boost Usage Quotas
As AI tools become more common, quota limits often block productivity. OpenAI recently announced a new Codex promotion on social media, offering tangible benefits to users.Running for two weeks, this campaign rewards users who invite friends to join
Meta launches AI optimism ad set to song about human extinction
Meta’s latest ad opens with a stark black-and-white close-up of an eye, capturing the anxiety of scrolling through alarming headlines about AI stealing jobs, isolating society, and triggering a global crisis.“Some people will have you believe AI is g
Mistral AI’s Open-Source Math Proof Tool Uses 119B Parameters, Activating Just 6B, Cuts Costs to 1% of Rivals
Mistral AI, a European artificial intelligence firm, has officially released Leanstral 1.5, a specialized model for formal mathematical proof. Built specifically for the Lean4 programming language, this 119B-parameter model activates only 6B paramete
Related Special Topic Recommendations
Comments (0)
0/500

Chinese large models are delivering a "dimensional strike" in the global computing power market through extreme cost-effectiveness.
According to the latest report, the Chinese large model MiniMax M2.5 has topped global large model usage rankings for five consecutive weeks. This achievement not only sets a new record for Chinese model activity on the international stage but also highlights China's core competitive advantage in the global AI industry: domestic models are priced dozens of times lower than overseas competitors with comparable capabilities.
This cost advantage, which leaves competitors struggling, stems from two strategic moats:
Underlying architectural innovation: making tokens more "valuable"
MiniMax developers note that fundamental innovations in the underlying architecture have significantly reduced inference costs. In simple terms, improved algorithm efficiency allows the model to complete the same complex tasks using fewer tokens, directly lowering users' computing costs at the technical level.
Energy and infrastructure advantages: building a computing power defense line
Beyond technological optimizations, China's energy advantage in computing power infrastructure cannot be overlooked. Industry data shows that electricity costs account for 70% to 80% of computing costs. For AI clusters running millions of graphics cards in parallel, China's stable energy supply and relatively low industrial electricity prices provide a highly competitive "cost firewall" for the AI industry.
From "computing power input" to "intellectual output," Chinese large models represented by MiniMax are rapidly capturing global market share through the dual forces of technological iteration and cost control. As AI competition enters its second half, this "high intelligence, low price" model may completely rewrite the rules of global large model commercialization.
OpenAI Unveils Codex Referral Program to Boost Usage Quotas
As AI tools become more common, quota limits often block productivity. OpenAI recently announced a new Codex promotion on social media, offering tangible benefits to users.Running for two weeks, this campaign rewards users who invite friends to join
Meta launches AI optimism ad set to song about human extinction
Meta’s latest ad opens with a stark black-and-white close-up of an eye, capturing the anxiety of scrolling through alarming headlines about AI stealing jobs, isolating society, and triggering a global crisis.“Some people will have you believe AI is g
Mistral AI’s Open-Source Math Proof Tool Uses 119B Parameters, Activating Just 6B, Cuts Costs to 1% of Rivals
Mistral AI, a European artificial intelligence firm, has officially released Leanstral 1.5, a specialized model for formal mathematical proof. Built specifically for the Lean4 programming language, this 119B-parameter model activates only 6B paramete











