User pins ChatGPT silent long-response truncation to server side after 3-client, dual-egress testing
jame-lee-cpu · ghdev · 2026-10-16
A user filed a detailed issue on openai/codex reporting that since Oct 16, longer Chinese ChatGPT responses repeatedly stop mid-generation — output cuts off while the UI presents it as a normally completed turn, with no network/quota error or notification; blank assistant responses were also observed.
The reporter systematically ruled out local causes:
- Reproduced on three clients: macOS native app (ChatGPT Classic 1.2026.184), macOS browser, and iOS
- Persisted across different proxies and two entirely different egress paths
- Client install, proxy software, egress IP, and OS network stack eliminated one by one
Conclusion: the fault is on ChatGPT's side — either the server ends the stream early or the client mishandles an interrupted/empty stream. The session model slug was gpt-6 (gpt-6-pro / gpt-6-thinking: max).
More from Models
- OpenAI blocks manual reset with 13% usage left: 'no reset needed right now' — AccomplishedSugar490 · 2026-10-23
- Meta's 816B Muse Spark Open scores 39, losing to 180B Qwen3.8-Flash-Next at 40 — BTA_Labs · 2026-10-23
- Gemini refuses to generate image in Docs, blames document size limit — viksit · 2026-10-23
- Dev jokes his true calling is telling everyone to try the latest DeepSeek — intellectronica · 2026-10-23
- SubQ Snap 1.1 ties for first on Artificial Analysis Decision Index with 384K context — ArtificialAnlys · 2026-10-23
- PolyAI unveils Dialog-RSN-1: audio-native voice model answering in under 300ms — matthen2 · 2026-10-23