OpenAI has launched Flex processing, a new API option designed to provide cheaper AI model usage by sacrificing response times and occasional unavailability of resources. This move aims to boost its competitiveness against AI giants like Google. Flex processing, which is currently in beta for OpenAI’s o3 and o4-mini reasoning models, targets lower-priority tasks such as model evaluations, data enrichment, and asynchronous workloads. The offering is designed to lower the cost of using these models, cutting API prices by 50%.
For the o3 model, Flex processing reduces input token prices to $5 per million (approximately 750,000 words) and output tokens to $20 per million, compared to the standard rates of $10 per million input tokens and $40 per million output tokens. The o4-mini model’s Flex pricing drops input token costs to $0.55 per million and output tokens to $2.20 per million, significantly cheaper than the regular rates of $1.10 and $4.40 per million.
This launch comes at a time when AI model prices are steadily rising, with competitors like Google offering more cost-effective options. Recently, Google introduced its Gemini 2.5 Flash reasoning model, which provides comparable or even superior performance to DeepSeek’s R1 at a lower input token cost.
Along with the introduction of Flex processing, OpenAI has implemented an ID verification requirement for users in the first three usage tiers to access the o3 model. This measure aims to prevent abuse of OpenAI’s services and ensure compliance with its usage policies.
