OpenAI reduces prices for its smaller models as companies evaluate their AI expenditures
OpenAI on Thursday, reduced the prices of its low- and mid-tier AI models, a decision that could heighten competition within the industry as U.S. companies strive to compete with more affordable Chinese counterparts for customers who are becoming increasingly cautious about the rising costs of technology.
The creator of ChatGPT has reduced the price of its smaller GPT-5.6 Luna model by 80% and its mid-tier Terra model by 20%, while maintaining the price of its largest and flagship Sol model.
The cuts indicate that the increasing scrutiny of costs by businesses dealing with substantial AI expenses is compelling American labs to reevaluate their pricing strategies. Recently, numerous tech CEOs have expressed that more affordable AI options are essential for the widespread adoption of the technology.
OpenAI’s new pricing intensifies competition with Anthropic, whose Claude models lead in enterprise and developer applications but are positioned at the higher end of the market. Both companies have faced challenges from open-source Chinese competitors like Z.ai’s GLM-5.2, which closely rival their performance while offering a more affordable option.
Analysts have indicated that reducing prices might enhance the adoption of OpenAI’s and Anthropic’s technology, though it could also put pressure on their finances in the lead-up to eagerly awaited initial public offerings.
While Thursday’s cuts impact only OpenAI’s smaller and mid-tier models, the company stated that it would still benefit businesses broadly, as those models can now perform tasks that previously required a top-tier system at a significantly lower cost.
The updated pricing structure indicates that businesses utilizing OpenAI’s technology will incur lower costs for each million “tokens,” which serve as the units for measuring AI usage, processed through the models.
Sending text to Luna decreases to 20 cents per million tokens from $1, and Terra’s drops to $2 from $2.50. Meanwhile, generating responses reduces to $1.20 and $12 from $6 and $15, respectively.
Anthropic’s mid-tier Claude Sonnet 4.6 model costs $3 per million input tokens and $15 per million output tokens, which is higher than the rates for Terra.
OpenAI stated that the reduced prices were, in part, facilitated by efficiency improvements from GPT-5.6, which include the model’s capability to enhance code and optimize performance during internal development.
In the past year, token prices have generally declined, while the expenses associated with completing tasks are increasing as AI companies transition from flat subscription models to usage-based pricing.
That results in companies facing unpredictable and frequently elevated bills as estimating usage per task becomes increasingly challenging.
