Users report GPT 5.6 Sol skipping reasoning on high mode and degrading output quality
Markuslanger25 · reddit · 2026-09-28
A Reddit user reports that since the start of the month, GPT 5.6 Sol in the Chat area frequently skips thinking entirely even on high reasoning: output either appears instantly or after 10-20 seconds and remains incomplete.
Key complaints:
- When reasoning is skipped, output style degrades into overly long, error-riddled formatting
- A zip-the-project analysis task that used to take over a minute of careful reading now finishes in 20 seconds without loading all files into context — the model itself admits it only skimmed
- The author finds Codex limits worse than Claude's €15 Opus 5.5 plan and says the Chat area is the only reason they subscribe
The post fuels ongoing user frustration about perceived quiet quality degradation in subscription models.
Related event: Users Report Quality Drop in GPT Sol Models(2 posts)→
More from Models
- Indie Dev From a Kerala Village Tops Hugging Face Trending With Laya Model — kalyan_kpl · 2026-09-28
- Jev Reads 384 News Items in 24.9s for $0.19, Beating Claude Opus 5 — FinanceYF5 · 2026-09-28
- Mystery model slug gpt-7-quetzalcoatl spotted on OpenAI, priced $0.05/$0.50 per M tokens — arthurcolle · 2026-09-28
- Gemini 4 Pro reportedly testing on Arena under the alias "Gemini 3.8 Flash" — DaniDFdesign · 2026-09-28
- Xiaomi MiMo-V2.6 fixes tool-call repetition with a 12-step RL teacher at ~4% retrain cost — LegacyRemaster · 2026-09-28
- Stanford runs OpenAI's GPT-6 Astra on a Unitree G1 robot to tidy an unseen kitchen — alex_verem · 2026-09-28