The author reports an issue with the Codex desktop app where usage is inaccurately attributed to a model (gpt-6-astra) that they did not select, leading to abnormal quota depletion. Despite only using the gpt-6.1 Sol model, the analytics chart shows that a significant portion of their usage (80%) is linked to gpt-6-astra. This discrepancy disrupts their work, and they've reached out to OpenAI support without receiving a response.