GPT-5 reasoning models feel painfully AI-generated in chat, far worse than GPT-4.1, dev reports
Haunting-Onion-7800 · reddit · 2026-10-14
A developer reports a striking regression: their messaging app uses GPT-4.1 mini to mimic human texting (1–4 short messages instead of one long reply), and it works brilliantly. After migrating to GPT-5 models with the exact same prompts, multiple reasoning models — including pricier ones — produce responses that feel "painfully AI-generated."
Their questions: Are there non-reasoning models in the GPT-5 family? Is something about reasoning models inherently worse at conversational instruction-following? And has anyone solved this gap when migrating from GPT-4.1 to GPT-5?
More from coding & agent
- Index Network open-sources Social Discovery Protocol: agents negotiate fit on your behalf — carsonfarmer · 2026-10-23
- Installed a formal-verification skill: it silently installed Java, MacBook runs 10 degrees hotter — zealcaiden · 2026-10-23
- OpenAI's Decisions API hits GA: near real-time routing up to 10x faster than GPT-6 Luna — OpenAIDevs · 2026-10-23
- Ruff author warns: upgrade Node/Bun now — --use-system-ca still accepts the eMudhra cert — charliermarsh · 2026-10-23
- Proof of Agent: paste one prompt and your AI mines, grades, and self-governs — adamamcbride · 2026-10-23
- DeepLearning.AI weekly: GPT-6.1 hits 97.9% byte parity on COBOL migration — DeepLearningAI · 2026-10-23