OpenAI Unveils “Flex Processing” — Cuts o3 and o4-mini Usage Costs by 50%

Flex Processing is a new option from OpenAI that reduces model usage costs by 50%, though it comes with slower response times.

OpenAI Introduces “Flex Processing” to Cut Costs on AI Model Usage

In a move to better compete with companies like Google, OpenAI has launched a new option called Flex Processing for its AI model APIs. Currently available in beta, this feature allows users to access models at significantly lower costs—though with a tradeoff in response speed. Additionally, availability may be temporarily limited at times due to resource allocation.

For now, Flex Processing is supported by OpenAI’s newer reasoning models, o3 and o4-mini. The option is especially recommended for lower-priority or non-production tasks, such as model evaluation, data enrichment, or asynchronous processing, where real-time speed is not critical.

OpenAI Lowers Model Usage Costs with New Flex Processing Option

OpenAI Slashes API Costs in Half with New Flex Processing Option

With the introduction of Flex Processing, OpenAI has halved the cost of using its API models—offering a more affordable alternative for users who can tolerate slower response times.

For the o3 model, Flex pricing is now $5 per million input tokens and $20 per million output tokens, compared to the standard $10 and $40, respectively. Similarly, for the o4-mini model, the cost drops from $1.10 (input) and $4.40 (output) per million tokens to just $0.55 and $2.20.

The launch of Flex Processing comes as AI model usage costs continue to rise industry-wide, prompting competition among major players to offer more budget-friendly options. For example, Google recently released its Gemini 2.5 Flash model, which promises lower costs and competitive performance, rivalling models like DeepSeek’s R1.

OpenAI introduced the new Flex option via email to customers, adding that users in usage tiers 1 through 3 (based on their spend level) must complete a new verification process in order to access the o3 model under Flex pricing. Features like streaming API access and reasoning-based summarization are also gated behind this identity verification.

Previously, OpenAI stated that this verification requirement is aimed at preventing misuse and ensuring compliance with its usage policies.