OpenAI Codex New Models Criticized for Fast Quota Burn and Task Failures
jdjohnson · x · 2026-07-22
User feedback indicates that OpenAI Codex's new models (Sol Medium and Terra Medium) perform poorly in practical office scenarios. Tests reveal that the models frequently make mistakes with basic tasks, such as generating a Scope of Work (SOW), and even fail to correctly set up files in Google Drive using defined skills. Furthermore, the models burn through usage limits rapidly: a user reported consuming 10% of their weekly quota just to generate two job descriptions and one SOW since the daily reset. The user suggests that normal users won't switch to Work/Codex unless the models become significantly more efficient or quotas are increased.
Related event: OpenAI Codex Faces Backlash Over Rapid Quota Drain and Poor Performance(3 posts)→
More from coding & agent
- Matt Pocock floats a skill naming scheme for agent writing workflows — mattpocockuk · 2026-07-23
- Live tutorial: how to build a Claude Code-style coding agent — dfinke · 2026-07-23
- Agent harnesses should be retested every time the model changes — hugobowne · 2026-07-23
- Shadow Web open-sources a local MCP browser stack with 20+ agent tools — Salt_Apartment5489 · 2026-07-23
- Claude voice mode now calls external tools, reaches email and Notion, and deploys to Vercel — koltregaskes · 2026-07-23
- Netlify pairs Drop with Agent Runners for drag-and-drop site edits in minutes — thisiskp_ · 2026-07-23