OpenAI's Codex Usage Instability Reflects Structural Struggle Between Growth and Compute Costs

This volatility is part of a broader transition in OpenAI's resource and pricing model. The company has introduced a new Pro 500 plan at $500 per month for its highest usage limits, while simultaneously reducing the allowance of the existing $200 Pro plan from 20 times to 10 times that of the Plus tier.

To manage compute demand, OpenAI has launched an "Ultrafast" mode that increases token generation speed for Enterprise and Pro 500 subscribers. This segmentation of performance and capacity suggests a strategy to optimize infrastructure costs as the company faces increasing demand for high-intensity agentic workflows.

Technical disputes over limit calculations have further eroded user trust. While OpenAI attributes abnormal usage drain to "sub2api"—an unsupported workaround to share subscriptions—developers argue the issue stems from a system bug that mistakenly applied weekly limits to a previous five-hour window. This follows a history of incidents where fraud-prevention systems incorrectly rate-limited legitimate accounts.