The landscape of artificial intelligence (AI) is undergoing a transformative shift, fueled by three critical forces that are reshaping the cost structure of AI technologies. Understanding these dynamics is essential for businesses and developers navigating the evolving AI ecosystem.
-
Foundation labs moving up the stack into applications:
Initially, foundation labs focused on developing the underlying models and architectures for AI, laying the groundwork for a broad range of applications. However, as competition intensifies, many are shifting their focus to higher-level applications. By doing so, they can offer more tailored solutions and services, thus creating greater value for their users. This transition not only affects the pricing of models but also opens avenues for innovative applications, as these labs leverage their foundational capabilities to offer specialized, industry-specific solutions.
-
Rising prices of frontier models:
The cost of accessing the most advanced AI models—often referred to as frontier models—continues to climb. As these models become more sophisticated, the investment required to train and implement them increases correspondingly. For enterprises and developers, this presents a significant challenge, as the high costs can deter the adoption of these cutting-edge technologies. The ongoing escalation in prices for frontier models forces organizations to reassess their strategies, exploring alternatives that maintain performance while reducing costs.
-
Open-source models crossing the ‘good enough’ threshold:
On the flip side, the rise of open-source models has provided a viable alternative for many businesses. These models have reached a level of performance that satisfies most use cases, allowing companies to harness powerful AI capabilities without incurring exorbitant fees. As open-source offerings improve, organizations increasingly find them sufficient for their needs, leading to a broader acceptance and integration of these models into various applications.
As AI buyers grapple with these shifting forces, a natural response has emerged: substitution. Businesses are actively seeking alternative models that align better with their budget constraints while maintaining or even enhancing performance levels.
Take Coinbase, for example:
“At Coinbase we’re working hot on routing prompts to cheaper models where appropriate, & in some cases have been able to keep costs roughly flat, while token usage continues to grow exponentially.”
This illustrates a proactive approach where organizations are optimizing their AI strategies, seeking to allocate resources effectively while managing costs.
Similarly, Lindy recently made a significant shift:
“Pulled the trigger today & switched 100% of Lindy traffic to DeepSeek v4, churning from Anthropic models. Saves us millions of $ & we’re actually seeing an increase in performance on many core use cases. Transformative for the business.”
This case showcases how making the right switch can deliver both cost savings and performance enhancements, underscoring the importance of adaptive strategies in the rapidly changing AI landscape.
Harvey also reported promising outcomes when testing AI models:
“On a 100-task slice of our Legal Agent Benchmark (LAB), SFT moved Kimi 2.6’s all-pass rate from 11% to 15%, beating Opus’ 14%. But the cost gap was even more striking: $84 vs $954 across the same 100 tasks, or ~11x cheaper.”
This example not only highlights significant cost differences but also emphasizes the need for continuous evaluation of model performance and pricing to stay competitive.
Even further, Cursor has taken innovation to the next level by post-training:
“Composer 2.5 is exceptionally intelligent & up to 10x more efficient than similarly capable models.”
This demonstrates how organizations can capitalize on their investments in AI by refining and customizing existing models for better efficiency in practical applications.
Coinbase’s proactive measures reflect a broader trend: as costs flatline while demand for tokens rises, companies often reinvest their savings into pursuing greater intelligence and capabilities. This cycle encourages a culture of continuous improvement and exploration of innovative solutions.
The current AI ecosystem presents a dichotomy: while closed models are becoming increasingly costly at the frontier, open models are concurrently becoming cheaper and rivaling proprietary technologies. The challenge for buyers lies in determining which economic slope best aligns with their business models and long-term goals.
