OpenAI, Meta and SpaceXAI race to cut enterprise AI costs

OpenAI, Meta Platforms and SpaceXAI have each unveiled new generative‑AI models that emphasise token efficiency and lower pricing for corporate users. OpenAI’s GPT‑5.6 model is promoted as up to 54 % more efficient on tasks such as code generation and uses fewer tokens per request, allowing enterprises to reduce AI spend. Meta’s Muse Spark 1.1 is billed as “very attractive” in price, with CEO Mark Zuckerberg framing the launch as an aggressive move toward cost‑effective services. SpaceXAI’s Grok 4.5 claims double the token‑usage efficiency of competing models, and the company is investing $9 billion in a new data‑centre in Canada to support the offering.

The push for cheaper AI comes as corporate finance teams scrutinise IA‑related invoices, a trend described by analysts as “token‑maxxing”. Open‑source providers such as Hugging Face are also gaining traction, with more than 30 % of Fortune 500 firms holding verified accounts, highlighting a broader market shift toward cost‑control and greater model transparency. Together, these developments illustrate an industry‑wide effort to balance advanced capabilities with tighter budgets, potentially reshaping enterprise AI adoption worldwide.