OpenAI's Chat Completions API may fail to report token usage when a streaming request is interrupted or cancelled. This can lead to inaccurate billing, as the missing usage is not recorded as zero or estimated locally. An open issue on the openai-openapi repository, filed in April 2026, details this problem and notes that OpenAI contributors have reproduced the bug. While potential solutions like running token counts on each chunk or retrieving usage post-completion are discussed, none are currently documented as reliable fixes for interrupted streams. AI
IMPACT Inaccurate billing for OpenAI API users could lead to financial discrepancies and require developers to implement custom workarounds for usage tracking.
RANK_REASON The cluster discusses a bug in a specific API product and its implications for users, rather than a new release or major industry shift.
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →