RU

ChatGPT 6.0: Personality and Depth Gone — Throttled Reasoning or a Model Swap?

Published: 2026-11-22 · Author: AI Release · @ai_releaseorg
ChatGPT 6.0: Personality and Depth Gone — Throttled Reasoning or a Model Swap?

⚡ The gist

In an OpenAI Community thread, user tamaradenmeade noticed that within 24 hours, the 6.0 model's answers got shorter and lost the "personality" she'd spent months building. With 78 replies and 102 likes (1,526 views), several other Plus subscribers chimed in reporting the same picture. The debate is over the cause: throttled reasoning, a quiet swap to a cheaper model, or a deliberate behavior change on the product side.

🗣 What participants are saying

Thread author tamaradenmeade describes a specific shift: her texts used to come out at "3000+ words," now they're "under 500–1000"; instead of a story, the model spits out a bare sequence of facts. The same thing happened in a new conversation inside a project, even though style instructions lived in the prompt, the source documents, and the custom instructions. She also flags the speed: the answer came back "instantly" instead of the usual wait.

A user named isha confirms: "personality was COMPLETELY stripped away," with the same 500 words in two or three paragraphs, no tone. He says he "greatly trained" his personality and it was all working just yesterday. He adds that he's on High thinking mode but sees no depth.

henri2 offers a casual test: the model used to respond to a comic with a joke and throw in analogies; now it "clinically explains mechanism and signs off." His framing of the conflict: "I don't want to use instant, I want thinking" — the option exists, but he can't feel any difference in reasoning.

skullofskelletor says the same thing from another angle: he spent a long time shaping a voice he could recognize from conversation to conversation, and now it "no longer sounds like itself." Even with High selected, he gets no depth beyond Instant, and the answers have become "radically shorter."

The sharpest argument from tamaradenmeade comes a day later: Instant behaves normally, while medium and high "act like a completely different model," at about a third of the usual length. She suspects she's been "dropped back" to free-tier level, even though her Plus subscription is active and there were no signs of a limit; in the end, she went back to 5.6, admitting it "doesn't have the same level of intelligence." In her final message, she pushes back on an explanation from another thread — that high responds the same way it does for work tasks: "if I choose CHAT, doesn't that mean I want a conversation?"

A developer named Phivo sees it differently: the new version "completely ignore established rules, wastes tokens" and can't handle tasks the previous model handled fine. maasalexandraj reacts emotionally: "It looks as if my Chat GPT got Alzheimer's" — and talks about losing trust in the tool.

⚖️ Where they agree, where they split

The common denominator is simple: several independent Plus users are seeing the same thing — shorter, flatter answers with no individual voice — and it all happened around the same time. Nobody in the thread disputes the observation itself; they only disagree on the explanation: a version swap, reasoning throttled under load, or a deliberate product shift toward a "workplace" tone.

The real conflict sits in two places. First, nobody has produced evidence: no screenshots with parameters, no comparison of the same prompt across different modes, no confirmation from OpenAI. Second, the accounts diverge on the facts: tamaradenmeade says Instant works as before while medium and high suffer; Phivo, on the contrary, complains about ballooning token usage, not compression. This isn't a consensus — it's a set of similar complaints from different scenarios: creative writing and development.

🧪 What to test yourself

🔗 Sources

OpenAI Community — 13 November/14 November AEDT Model changes to 6.0 in the last 24 hours or has reasoning been throttled

Disclaimer: this is a recap of a discussion, not facts verified by our editors. The opinions belong to the participants; the numbers belong to whoever cited them.

🤖 AI summary
  • Date: 2026-11-22
  • Source: community.openai.com
  • Tags: AI, ChatGPT, reasoning
  • Key facts:
    • In an OpenAI Community thread, user tamaradenmeade noticed that within 24 hours, the 6.0 model's answers got shorter and lost the "personality" she'd spent months building.
    • With 78 replies and 102 likes (1,526 views), several other Plus subscribers chimed in reporting the same picture.
    • The debate is over the cause: throttled reasoning, a quiet swap to a cheaper model, or a deliberate behavior change on the product side.
    • Thread author tamaradenmeade describes a specific shift: her texts used to come out at "3000+ words," now they're "under 500–1000"; instead of a story, the model spits out a bare sequence of facts.
    • The same thing happened in a new conversation inside a project, even though style instructions lived in the prompt, the source documents, and the custom instructions.
    • She also flags the speed: the answer came back "instantly" instead of the usual wait.
    • A user named isha confirms: "personality was COMPLETELY stripped away," with the same 500 words in two or three paragraphs, no tone.
    • He says he "greatly trained" his personality and it was all working just yesterday.
AIChatGPTreasoning
← PreviousSignal blocks AI screenshots on iOS and Android: a response to Gemini and Siri

Source: community.openai.com · post in Telegram