OpenAI Cuts Smaller AI Model Prices As Companies Push Back On AI Costs

OpenAI has taken the dramatic step of slashing prices for its smaller and mid-tier AI models, a clear signal that the AI industry is entering a new phase of cost competition.

The company announced that it is cutting the price of its GPT-5.6 Luna model by a staggering 80% and its mid-tier Terra model by 20%. This aggressive move comes just three weeks after the models were launched and reflects a growing pushback from businesses grappling with the ballooning costs of AI deployment.

The price cuts are part of a broader trend where the focus is shifting from pure capability to cost efficiency, as companies seek to make AI a sustainable part of their operations.

Under the new pricing structure, sending text to Luna now costs just 20 cents per million input tokens, down from $1, while output drops to $1.20 from $6. Terra's pricing has been adjusted to $2 for input and $12 for output, down from $2.50 and $15, respectively.

The flagship Sol model remains unchanged at $5 for input and $30 for output per million tokens. This strategic pricing makes OpenAI's models significantly more affordable, directly countering the pressure from open-source Chinese rivals like Z.ai's GLM-5.2, which nearly match the performance of frontier models at a much lower cost.

"The cuts show that rising cost scrutiny by businesses facing hefty AI bills is forcing American labs to rethink pricing. Many tech CEOs have also said in recent months that cheaper AI options are key to the technology's widespread adoption." – Reuters analysis

The price reductions are a direct challenge to Anthropic, whose Claude models are popular in the enterprise but are now more expensive than OpenAI's offerings. For instance, Anthropic's mid-tier Claude Sonnet 4.6 costs $3 per million input tokens and $15 per million output tokens, which is above the new rates for Terra.

This move intensifies the competitive landscape, pushing other AI providers to reconsider their pricing strategies and forcing businesses to reevaluate which models to use for different tasks.

The AI cost efficiency improvements that enabled these cuts were partly driven by the models themselves. OpenAI stated that GPT-5.6 Sol helped optimize production GPU kernels, reducing inference costs without compromising performance.

Overall efficiency gains reduced the end-to-end cost of serving the model by 20% and increased token-generation efficiency by more than 15%. For businesses, the impact of these price cuts extends beyond just lowering bills.

Analysts suggest that CIOs are likely to reinvest savings into higher AI usage, enabling more complex agentic workflows that were previously difficult to justify economically. This reflects the Jevons paradox, where increased efficiency leads to greater overall consumption, ultimately driving more value from AI investments.