OpenAI has announced a new API pricing option called Flex processing, aimed at developers and organizations seeking more affordable access to its AI models. Flex offers a lower-cost alternative to standard processing by allowing for slower responses and the possibility of occasional unavailability. The new feature is now available in beta for the company’s o3 and o4-mini reasoning models.
Flex processing targets low-priority or non-essential tasks such as model evaluation, data enrichment, and asynchronous operations. By opting for Flex, users can reduce their API expenses by half. For the o3 model, pricing drops to $5 per million input tokens and $20 per million output tokens—down from $10 and $40, respectively. The o4-mini model sees similar reductions, with input tokens costing $0.55 per million and output tokens priced at $2.20 per million, compared to standard rates of $1.10 and $4.40.
This cost-cutting option arrives as the overall expenses of high-end AI usage continue to grow. At the same time, competitors like Google are pushing out cheaper, performance-optimized models. Just this week, Google launched Gemini 2.5 Flash, a reasoning model designed to offer strong performance at reduced input token costs, positioning it against other efficient models like DeepSeek’s R1.
Alongside the Flex rollout, OpenAI informed users that those in usage tiers 1 through 3—determined by their spending level—must complete a new ID verification process to access o3 and other advanced model features. This includes capabilities like reasoning summaries and streaming API support. The company says this verification step is intended to help prevent misuse and ensure that its services are used in line with policy guidelines.
By introducing Flex processing, OpenAI aims to give developers more flexibility in managing costs, especially for tasks that don’t require immediate or high-priority results. The move also strengthens its position in the competitive AI space, offering a pricing structure that could appeal to startups, researchers, and businesses with budget-conscious AI needs.
