OpenAI Codex Faces Backlash Over Rapid Credit Drain and Model Errors

Recently, multiple developers took to social media to report severe credit drains and poor model performance with OpenAI Codex, raising concerns about its commercial viability.

Confirmed

Several users reported rapid credit consumption. Users @willdepue and @jxnlco noted that about $400 in credits were exhausted in just 4 hours while running approximately 4 high-priority agents, suspecting background loop tasks were responsible. User @jdjohnson found that after a credit reset, using only the Sol Medium and Terra Medium models to complete two job descriptions and one scope of work document consumed 10% of the weekly quota. Additionally, @Acceptable-War4836 pointed out that while Plus subscription web chat is nearly unlimited, Codex drains a massive amount of the weekly quota in fewer than 20 prompts. Beyond credit issues, @jdjohnson highlighted that the new models frequently fail in practical office scenarios, making errors on basic tasks like generating an SOW and failing to correctly invoke skills to read Google Drive.

Why it matters

Codex is positioned for heavy development and office scenarios, but the current rate of credit consumption and model instability clearly cannot support a full transition for regular users. If OpenAI does not optimize its billing mechanism (especially the consumption logic for concurrent agents) and underlying model capabilities, it will severely dampen developer willingness to use the platform.

2026-07-22 ~ 2026-07-24 · 5 related posts

Primary sources