Wave Goodbye to Tokenmaxxing, Modelmaxxing Is Here to Stay
In the rapidly evolving world of artificial intelligence, there's a new trend catching on - modelmaxxing. Tech leaders like Morgan Linton, a top-gun in an AI startup based in Lake Tahoe, are leading the charge. Linton is very specific about which AI models his engineering team uses and when, ensuring they get the most out of each model without over-burdening them.
"This approach allows for more efficient use of AI, and my team gets to work with the best tools in the business," Linton explained.
In the first half of 2026, a trend known as tokenmaxxing dominated the AI landscape. This involved companies encouraging their employees to use as much AI as possible. However, after evaluating the hefty bills racked up by excessive AI usage, corporations from various industries have started to rethink their approach.
Model Switching: The Cost-saving Hack
Everyone from founders, software engineers, UX designers, and even non-technical enthusiasts are adopting a clever cost-saving measure: model switching. This involves assigning complex tasks to more expensive cutting-edge models and delegating simpler, repetitive tasks to older, cheaper models. With businesses tightening their AI budgets and setting usage limits, this strategy could ensure you get the most out of every dollar spent.
Of course, there are advantages to using the most recent models. Kaylin Voss, an expert in the AI field, pointed out that superior models can reduce retries, supervision, and wasted effort. However, it's important to consider cost-effectiveness. As Brian Armstrong, CEO of a leading digital currency platform, stated, "80% of workloads will be running on 99% cheaper models within 12-18 months."
Chris Maconi, a leading entrepreneur in the AI startup scene, is also not a fan of tokenmaxxing. He believes in a "human-in-the-loop" approach and doesn't rely on bots for round-the-clock coding. Instead, he carefully selects his models, even going for lower-end models if they can deliver the needed intelligence.
Maximizing Model Efficiency
Tech professionals are also finding ways to stretch their tokens creatively. User-experience designer Tanvi Pisal, for example, learned the hard way to use models more efficiently. Using a design-first process, she saves tokens by brainstorming ideas with one AI tool and then using another to create more polished documents.
Alejandra Thomas, a software engineer based in New York City, tests each new model to determine its strengths. "For simpler tasks, I always use lighter models or none at all," she said. Similarly, Ed Stevens, CEO of an AI sales company, believes in testing a model for a few months before deciding whether it's worth the investment.
Dan Ariely, a behavioral economics researcher, linked this mindset to the days when cellphone plans had limited talk time. "Tokens create a model of scarcity where people can't use as much as they want. Users switch to other models to save on costs once they've reached their token limit," he explained.
Model Routing Startups to the Rescue
If all this talk of modelmaxxing sounds tiring, don't worry. Model routing startups are stepping in to help. These companies provide software that assigns tasks to specific models based on complexity. These startups are gaining popularity and receiving hefty funding.
David Gilmore, who runs one such company, Rayline, shared that his tool determines if requests could go to cheaper models. He noticed many clients fall into a trap of wanting to use the latest AI models, only to realize they need to scale back after receiving their API bill.
The use of a routing platform is slowly but surely on the rise. Last year, only about 1% of firms used a model router, but this year, that number has risen to 5%.
However, despite the growing trend of modelmaxxing, some still default to the latest and most expensive models. As Chris Maconi noted, "People don't want to do the hard work of understanding which models are good at which things. They just want to ride the hype train."
In the rapidly evolving world of artificial intelligence, there's a new trend catching on - modelmaxxing. Tech leaders like Morgan Linton, a top-gun in an AI startup based in Lake Tahoe, are leading the charge. Linton is very specific about which AI models his engineering team uses and when, ensuring they get the most out of each model without over-burdening them.
"This approach allows for more efficient use of AI, and my team gets to work with the best tools in the business," Linton explained.
In the first half of 2026, a trend known as tokenmaxxing dominated the AI landscape. This involved companies encouraging their employees to use as much AI as possible. However, after evaluating the hefty bills racked up by excessive AI usage, corporations from various industries have started to rethink their approach.
Model Switching: The Cost-saving Hack
Everyone from founders, software engineers, UX designers, and even non-technical enthusiasts are adopting a clever cost-saving measure: model switching. This involves assigning complex tasks to more expensive cutting-edge models and delegating simpler, repetitive tasks to older, cheaper models. With businesses tightening their AI budgets and setting usage limits, this strategy could ensure you get the most out of every dollar spent.
Of course, there are advantages to using the most recent models. Kaylin Voss, an expert in the AI field, pointed out that superior models can reduce retries, supervision, and wasted effort. However, it's important to consider cost-effectiveness. As Brian Armstrong, CEO of a leading digital currency platform, stated, "80% of workloads will be running on 99% cheaper models within 12-18 months."
Chris Maconi, a leading entrepreneur in the AI startup scene, is also not a fan of tokenmaxxing. He believes in a "human-in-the-loop" approach and doesn't rely on bots for round-the-clock coding. Instead, he carefully selects his models, even going for lower-end models if they can deliver the needed intelligence.
Maximizing Model Efficiency
Tech professionals are also finding ways to stretch their tokens creatively. User-experience designer Tanvi Pisal, for example, learned the hard way to use models more efficiently. Using a design-first process, she saves tokens by brainstorming ideas with one AI tool and then using another to create more polished documents.
Alejandra Thomas, a software engineer based in New York City, tests each new model to determine its strengths. "For simpler tasks, I always use lighter models or none at all," she said. Similarly, Ed Stevens, CEO of an AI sales company, believes in testing a model for a few months before deciding whether it's worth the investment.
Dan Ariely, a behavioral economics researcher, linked this mindset to the days when cellphone plans had limited talk time. "Tokens create a model of scarcity where people can't use as much as they want. Users switch to other models to save on costs once they've reached their token limit," he explained.
Model Routing Startups to the Rescue
If all this talk of modelmaxxing sounds tiring, don't worry. Model routing startups are stepping in to help. These companies provide software that assigns tasks to specific models based on complexity. These startups are gaining popularity and receiving hefty funding.
David Gilmore, who runs one such company, Rayline, shared that his tool determines if requests could go to cheaper models. He noticed many clients fall into a trap of wanting to use the latest AI models, only to realize they need to scale back after receiving their API bill.
The use of a routing platform is slowly but surely on the rise. Last year, only about 1% of firms used a model router, but this year, that number has risen to 5%.
However, despite the growing trend of modelmaxxing, some still default to the latest and most expensive models. As Chris Maconi noted, "People don't want to do the hard work of understanding which models are good at which things. They just want to ride the hype train."