Companies are abandoning costly “tokenmaxxing” and turning to model routing, using cheaper AI systems for routine work while reserving advanced models for complex tasks, as executives question whether rising token use is producing measurable productivity, stronger products, or company-wide returns.
Usually referred to as “modelmaxxing” or “tokenmaxxing,” the shift directs each task to the model best suited to complete it at a lower cost, benefiting Chinese and open-weight developers while pressuring American OpenAI, Anthropic, and other premium providers, to justify higher prices.
From Token Leaderboards to Cost Controls
Tokenmaxxing gained attention as employers encouraged workers to use more AI and treated token consumption as innovation. Tokens are the units AI systems process and charge for, making usage easy to track but expensive to scale.
Meta, Amazon, and OpenAI reportedly used leaderboards to compare activity under tokenmaxxing. Higher totals were treated as evidence that workers were deploying AI agents, despite rising enterprise AI agent implementation costs.
The cost barriers of AI adoption in small and medium enterprises (SMEs) metrics soon created problems. Some Amazon employees reportedly assigned unnecessary work to AI agents after managers began considering usage.
The cost management strategies for AI behavior reflect Goodhart’s Law: when a measure becomes a target, it can stop measuring its intended purpose.
Costs also rose, turning tokenmaxxing into a warning about cost barriers of AI adoption in small and medium enterprises for smaller businesses.
Meta removed an informal token leaderboard, while Microsoft reportedly cancelled Claude Code subscriptions in several divisions. In parallel, Uber said it used its entire 2026 token budget during the first four months, partly because of heavy Claude Code usage.
Salesforce chief executive Marc Benioff said its Anthropic bill could reach $300 million this year and called for a smart router that determines whether a request needs a frontier model or cheaper alternative. Such tools could offer the best cost-saving strategies for AI orchestration.
OpenAI chief executive Sam Altman acknowledged the pressure at the Allen & Co. Sun Valley Conference. “This is the first year where AI spend has been a big topic,” OpenAI CEO, Sam Altman, told CNBC. “And all of a sudden, it’s a very big topic. Everyone’s asking what we can do to help reduce spend or increase value.”
These cases are pushing companies toward cost management strategies for AI deployment.
AI Model Routing Opens the Door to Chinese AI
AI model routing gives companies greater cost control.
A capable model routing AI can plan an assignment, while smaller models handle routine execution, making itpart of generative AI cost management strategies for business.
“The AI race is no longer just about building the best model,” said Soumen Mandal, principal analyst at Counterpoint Research.
“While performance will remain important for both enterprises and consumers, the right balance of capability, cost and deployment flexibility will ultimately determine the winners in the AI market.”
Chinese models are central to that cost management strategies for AI calculation. DeepSeek, Doubao, GLM, Hunyuan, Kimi, and Qwen have improved in coding, mathematics, multilingual work, and productivity. Moonshot’s Kimi K3 performed strongly, while Alibaba previewed Qwen3.8 Max.
According to an IDC survey, 47% of 260 US decision-makers at companies with more than 1,000 employees used Chinese-made models for at least one purpose, while 20% reported extensive use. Another survey found 73% used or tested AI-driven routing, and 72% tested automated routing.
Price is a major advantage. Mandal said DeepSeek and Qwen generally cost less than $5 per million output tokens, compared with roughly $25 to $30 for leading US models. Open-weight systems offer faster deployment and greater control, although closed models may provide stronger accuracy. This gap highlights the cost barriers of AI adoption in small and medium enterprises.
The also shows the need for cost management strategies for AI deployment.
Security and compliance remain obstacles, as companies must consider cost barriers of AI adoption in small and medium enterprises hosting, data processing, and whether sensitive information could enter untrusted infrastructure. Open-source derivatives can hide their architecture, creating procurement and governance risks.
Companies are building multi-model systems, balancing performance, cost, privacy, and reliability, which could reduce the cost barriers of AI adoption in small and medium enterprises.
Inside Telecom provides you with an extensive list of content covering all aspects of the tech industry. Keep an eye on our Intelligent Tech sections to stay informed and up-to-date with our daily articles.
