2 min read
OpenAI has unveiled a new pricing tier called Flex processing, designed for developers seeking to cut costs on AI tasks that aren’t time-sensitive. With this move, the company is aiming to provide a more affordable option for background or experimental work, while subtly tightening its grip on premium model access.
Available in beta for OpenAI’s latest o3 and o4-mini reasoning models, Flex offers up to 50% cost reduction compared to standard usage. For example, developers using the o3 model will pay $5 per million input tokens and $20 per million output tokens—half the regular price. The o4-mini model sees a similar cut, with Flex pricing set at $0.55 and $2.20 per million input and output tokens, respectively.
But the lower price tag comes with trade-offs.
Flex jobs are slower, occasionally delayed, and more prone to hitting resource limits. OpenAI warns developers of potential timeouts (default of 10 minutes) or 429 errors, recommending longer wait settings and retry strategies. That means Flex is better suited for non-urgent tasks, like model evaluation, data enrichment, or other async processes, not real-time applications.
Developers can enable Flex by simply adding a service_tier=”flex” parameter to their API calls.
Beyond pricing, OpenAI is also tightening access controls. Developers in lower spending tiers (1–3) must now complete ID verification to unlock full access to the o3 model and advanced features like reasoning summaries and streaming API support. These moves indicate OpenAI’s broader strategy: balance affordability for developers at scale while keeping premium tools more exclusive.
The announcement comes as AI rivals like Google step up their game, most recently with the release of Gemini 2.5 Flash, a fast, cost-efficient reasoning model.
With Flex, OpenAI is signaling it won’t just compete on performance, but on price flexibility and strategic access too.
