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 Credit Drain and Model Errors(5 posts)→
More from coding & agent
- Codex ban update: users jailbreak AI agents to bypass Daybreak filter for reverse engineering — FlolightC · 2026-10-27
- GPT-6.1 Terra hits GA in GitHub Copilot, stronger terminal coding at lower cost than GPT-6.1 Sol — 0xkarasy · 2026-10-27
- MIT: LM harnesses are compositional generalizers, not just scaffolding — akbirthko · 2026-10-27
- InterviewPlayground: JHU's simulated eval for AI interviewers hits 0.86 correlation — anjalie_f · 2026-10-27
- IDA 9.5 ships Qualcomm Hexagon decompiler plus IDA MCP for AI-driven reverse engineering — OwariDa · 2026-10-27
- Redditor builds a Star Trek-style home AI: the model was easy, the audio plumbing wasn't — FF2183 · 2026-10-27