Codex (GPT-5.5, Plus plan) – rate-limit cost per token jumped 10x+ since June 16 (github.com)

Users of the Codex (GPT-5.5, Plus plan) have reported a substantial increase in the rate-limit cost per token since June 16, with the budget for usage draining 10 to 20 times faster than previously experienced. Affected users found that their prompt count significantly dropped from over 20 to just 2-3 prompts, despite using smaller input sizes and recording minimal reasoning tokens. This discrepancy suggests that the per-token consumption has drastically changed without any user configuration adjustments. The significance of this development lies in its implications for accessibility and usability of the GPT-5.5 model in practical applications. If the rate-limit budget has intentionally been reduced or the weight per token adjusted, it could lead to increased costs for users and potentially stifle the creative and experimental opportunities that the AI offers. As users seek clarification from OpenAI regarding these changes, the incident highlights ongoing issues around model performance metrics and user experience that the AI/ML community must monitor closely.