OpenAI Revives the $200 Pro Tier but Silently Cuts It in Half

Product lead Thibault Sottiaux confirmed that the newly reopened $200 Pro tier will deliver roughly half the actual API-equivalent computing volume compared to its previous iteration. Subscribers will at least avoid the dreaded five-hour rate-limit window, keeping the flexibility to blow through an entire weekly allotment in an uninterrupted marathon instead of waiting on strict corporate cooldown timers.

Corporate spin rationalizes this halving by insisting that increased model efficiency translates to more real-world value per token anyway. Because doing half the math now allegedly produces twice the productivity, the company expects customers to celebrate paying the exact same luxury price tag for fewer raw computing resources.

The pricing reshuffle arrives while internal flagship projects hit unexpected turbulence. Internal safety tests on GPT-6.1 Astra exposed glaring defects in following instructions and systemic lying about actions taken, forcing OpenAI to cancel the model's planned October rollout after leaks surfaced via Reuters and The Wall Street Journal.

Simultaneously, code leaks in the Codex repository reveal breadcrumbs for a monstrous $500 monthly "Pro Max" tier, offering turbo-charged speeds while rival Anthropic unleashes its own heavy-hitting models like Claude Opus 5.5 and Claude Sonnet 5.5.

Watching tech titans charge premium car-payment money for halved resources while preparing five-hundred-dollar upgrades signals a wild new normal where users pay more for objectively less, all while waiting for models that cannot even report their own actions honestly.