#Bug Title ChatGPT gives inconsistent explanations about models, reasoning modes, routing, and tools

1 messages · Page 1 of 1 (latest)

obsidian trenchBOT
#

Reported by @west lynx

Bug Report: Bug Title ChatGPT gives inconsistent explanations about models, reasoning modes, routing, and tools
`Steps to Reproduce`

Steps to Reproduce

  1. Ask ChatGPT to explain:
  • Instant vs Thinking vs Pro
  • Standard vs Extended vs Heavy
  • Search behavior
  • Atlas/page-context behavior
  • Codex vs ChatGPT differences
  • routing/context behavior

  1. Compare answers across different GPT-5.5 modes.

  2. Compare those answers against:

  • official OpenAI docs;
  • UI behavior;
  • actual observed behavior.

Observed:
different models sometimes explain the same feature differently or incorrectly.

`Expected Result`

Expected Result

ChatGPT should reliably understand and explain:

  • model modes;
  • reasoning effort;
  • routing behavior;
  • Search;
  • Atlas/page context;
  • Codex differences;
  • and context systems.

If behavior is uncertain, the assistant should clearly say:

  • “not confirmed in official docs”;
  • “observed behavior”;
  • or “needs verification.”
`Actual Result`

Actual Result

ChatGPT sometimes gives conflicting or incorrect explanations about its own products and behavior.

Examples observed:

  • confusion between Atlas side chat and chatgpt.com web-app mode;
  • inconsistent explanations of reasoning-effort modes;
  • uncertainty around Search vs active-page context;
  • unclear routing explanations;
  • different GPT-5.5 modes explaining the same feature differently.

This can force users to experimentally reverse-engineer the platform instead of relying on t

`Environment`

Environment ChatGPT Pro GPT-5.5 Instant / Thinking / Pro Standard / Extended / Heavy testing Atlas browser Projects Web Search ON/OFF testing macOS