Users Report Sudden Quality Degradation in ChatGPT and Sol Models, Suspecting Routing to Smaller Model

Between August 21 and 22, multiple independent Reddit users reported sudden quality degradation in ChatGPT and OpenAI's Sol models across chat, Codex, instruction-following, and memory. Some users observed that the high-intelligence tier appeared to be routed to a smaller 5.5-mini model, responding much faster but with shallower reasoning. The cause remains unknown and OpenAI has issued no incident notice, while paying Plus users report real workflow impact.

Confirmed

• @Illustrious-Bet-1368 reported that Sol High (Chat and Codex) lost its reasoning chains, replying instantly but without thought, with degraded output and even profanity.
• @eesnimi relayed user reports of strawmanning user positions, forced rebuttals without understanding, and imposed frameworks within the past 12-24 hours.
• @panzzersoldat observed the "High" tier, which should call the 5.6 model, seemingly routed to 5.5-mini — instant-speed responses with shallow output — suspecting a server-side routing issue.
• @Middle-Wealth-6755, a Plus user on the 5.6 Sol model, reported a month of severe memory degradation, forgetting explicitly stated key information and ignoring instructions, and is collecting similar reports.
• @M-M103 said multiple GPT Plus users found GPT-4.1 Sol answers became random, irrelevant, and low-quality within hours.
• @Jacuzzi0317 reported basic errors and declining instruction-following over the past month; one user needed 10+ corrections on simple instructions even at maximum thinking mode, feeling like a "20-30% IQ drop."
• @XGatsbyX called the recent update a "disaster": inconsistent multi-platform UI, chaotic Codex integration, lost sessions, uncontrollable model/thinking-mode switching, and replies too degraded to scan codebases or grasp context.
• @markk found GPT-4.1 Sol dramatically faster after a month away, suspecting optimization or downgrade, though generation quality seemed acceptable in their experience.

Unconfirmed

• @M-M103 noted it is unclear whether this is a broad service failure or A/B testing targeting some users.
• The "routed to 5.5-mini" claim is @panzzersoldat's personal observation; @markk's "optimization or downgrade" guess is likewise unverified, and their acceptable quality contrasts with other reports.
• All degradation reports are subjective user experiences; OpenAI has published no incident notice or explanation.

Why it matters

• Many independent users across subreddits reported similar degradation concentrated around August 21, suggesting a possible backend model or deployment change rather than isolated cases; the "smaller model routing" and "sudden speed increase" observations corroborate each other.
• If true, it affects paying Plus users' daily productivity tools, with declines in memory, instruction-following, and code comprehension directly impacting workflows as users seek workarounds.

2026-08-21 ~ 2026-08-22 · 9 related posts

• Users report GPT-4.1 Sol model suddenly became dumb with irrelevant answers — M-M103 · 2026-08-21
• Users Report Significant Performance Degradation in ChatGPT — Jacuzzi0317 · 2026-08-21
• ChatGPT Update Bashed: Broken UI, Integration, and Quality Drop — XGatsbyX · 2026-08-21
• Users report severe regression in ChatGPT behavior and reasoning — eesnimi · 2026-08-21
• [source] ChatGPT users report 'high' intelligence setting silently routing to 5.5-mini — panzzersoldat · 2026-08-21
• [source] Users report ChatGPT's memory has noticeably degraded over the past month — Middle-Wealth-6755 · 2026-08-21
• [source] Users Report Degradation in OpenAI Sol High Chat/Codex — Illustrious-Bet-1368 · 2026-08-22
• Users Report Significantly Faster GPT-4.1 Sol Response in ChatGPT — mark_k · 2026-08-22
• Users observe suspicious speed improvements in ChatGPT — mark_k · 2026-08-22