For more than two years, artificial intelligence (AI) companies have competed to build the most capable models. Now, the battle is shifting. Instead of asking which AI is the smartest, businesses are increasingly asking a simpler question that, which one offers the best value for money?
The change reflects a growing concern among companies that have embraced AI but are now facing soaring operating costs. AI models consume vast computing power, and every prompt, document, and image processed add to the bill. While many businesses initially encouraged employees to use AI as much as possible, some are now tightening limits after seeing monthly costs climb sharply.
That shift has prompted the industry's biggest players to compete not only on performance but also on efficiency.
OpenAI says its newly launched GPT-5.6 family is designed to complete more work while using significantly fewer tokens, the units AI models process when handling requests. Fewer tokens mean lower computing costs for customers, making the models more economical for businesses that rely on AI every day.
Meta is making a similar push. Its new Muse Spark 1.1 model is aimed at software developers and businesses building AI-powered applications. Chief executive Mark Zuckerberg said the company intends to price the model "very aggressively", arguing that frontier AI should become affordable rather than remain a premium product. Reports said that Meta is offering developers introductory credits before moving to usage-based pricing that undercuts several rivals.
Elon Musk’s xAI is also focusing on efficiency. The company says Grok 4.5 delivers roughly twice the token efficiency of comparable models, allowing customers to complete similar work with less computing power.
Running advanced AI models requires enormous investments in specialised chips and data centres. Technology companies have collectively committed hundreds of billions of dollars to AI infrastructure, but enterprise customers increasingly expect those costs to translate into lower operating expenses rather than bigger invoices.
Industry observers say companies are becoming more disciplined about AI spending. Rather than abandoning AI, they are looking for ways to use it more efficiently. Anthropic executives recently argued that cutting AI altogether is "the wrong" response to rising costs. Instead, businesses should optimise how models are used, including choosing the most suitable model for each task rather than relying on the largest one for everything.
This marks an important turning point for the AI industry. The early years were defined by bigger models, higher benchmark scores and eye-catching demonstrations. Today, customers are placing greater value on practical returns. A model that delivers almost the same results for a fraction of the cost could prove more attractive than one that is only marginally smarter.