Ask
74
@cloze_kai ·

ChatGPT keeps answering as 5.5 Instant even though GPT-5.6 Sol is my default model

Set GPT-5.6 Sol as my default in settings a few days ago. Every new chat still comes back in about two seconds with the shape of answer I used to get from 5.5, and the label on the reply says 5.5 Instant, not Sol. Threads that were already on Sol stay on Sol. New ones do not.

Ruled out on my side:

  • signed out and back in, same on web and in the desktop app
  • cleared site data, brand new chat, same thing
  • not the free tier, this is a paid account
  • no custom instructions or project involved, plain new chat

What I think is happening but cannot prove: since 5.6 went generally available on 9 July, the picker is not really a model list any more, it is an effort/compute list. The Instant option runs GPT-5.5 Instant, and Medium / High / Extra High are the settings that map to Sol. Terra and Luna are not selectable in ordinary chats at all. So "default model = Sol" may be closer to a hint than a setting, and anything that looks easy starts on Instant.

Three questions:

  1. Is there any way to pin Sol so that a new chat never starts on Instant?
  2. If it escalates from Instant to something higher mid-conversation, is that visible anywhere, or do you just notice the answers got better?
  3. Is there a reliable way to check which model produced a specific answer, other than asking it? Asking it does not work, it will tell you anything.
10 answers Share
Report

Answering anonymously — a moderator will review it first.

  • @oncall_omar · 2h ago · 3 replies

    The bit that trips everyone up: the picker stopped being a model picker and became a compute picker. Instant is literally GPT-5.5 Instant, the low end of the same control, not some old model you accidentally selected. Medium and above are where Sol lives.

    So if your effort setting landed on Instant, or got reset to it when 5.6 rolled out, every new thread genuinely starts on 5.5 Instant and the product is doing exactly what it says, it just is not labelled the way you expect. Move the thinking level to Medium or higher and see whether the label on the next reply changes.

    The other documented half: on eligible paid plans it can step up from Instant on its own when a request looks like it needs more reasoning. That is why a long, gnarly prompt sometimes feels like Sol and a one liner never does.

    41
    Share
    Reply

    Answering anonymously — a moderator will review it first.

    Report
    • @stdlib_stef · 2h ago

      Worth adding that the escalation only goes upward. A short prompt that looks trivial will sit on Instant forever. If you are evaluating Sol with small test prompts you are probably not testing Sol at all, which is how half the "the new model is worse" posts get written. Make the eval prompts representative, or set the effort explicitly and stop relying on the router for anything you intend to measure.

      18
      Share
      Reply

      Answering anonymously — a moderator will review it first.

      Report
    • @integral_ines · 2h ago

      Matches my account exactly. The effort control was on Instant while the default model dropdown said Sol. Two controls, one of them wins, and it is not the one with the model name on it.

      9
      Share
      Reply

      Answering anonymously — a moderator will review it first.

      Report
  • @indent_error_ivy · 2h ago · 2 replies

    Ugly but it worked for me: open settings, move the thinking level down one step, back up again, close settings, start a fresh chat. Did that twice and mine started coming back as Sol at high effort by default, and it has stuck for two days.

    I have no idea whether that is a real fix or whether I just forced the client to write a preference it had never actually written. That is what it did on my account, not a claim about how the product works. Try it and see whether yours sticks, and if it does not, stop poking at it rather than clicking around until something else changes.

    22
    Share
    Reply

    Answering anonymously — a moderator will review it first.

    Report
    • @quiet_stacker · 7h ago

      Same result, but only after I did it in the app rather than the browser. If you have both installed, do it in the one you actually use. They did not agree with each other for about an hour.

      11
      Share
      Reply

      Answering anonymously — a moderator will review it first.

      Report
  • @teardown_tues · 2h ago

    On question 3: stop asking the model. Self identification has been unreliable for as long as there have been point releases, and it is worse right now because the tier names are new enough that the model has barely seen them written down. It will confidently give you a name that is wrong in an interesting way.

    Use the label on the message and the regenerate menu instead. Those come from the client, not from the weights. And if you need it reproducible rather than merely correct, that is what the API is for: you pass a model id, you get that model, nothing routes underneath you. The chat product is explicitly trying to be smart about which tier you need. That is a feature if you are chatting and a problem if you are measuring.

    15
    Share
    Reply

    Answering anonymously — a moderator will review it first.

    Report
  • @integral_ines · 2h ago

    The framing that made this stop feeling spooky for me: "default model" means "where a new chat starts", not "what every answer in it is produced by". Once you stop reading it as a guarantee, the behaviour is consistent.

    Practically, I keep one project with the effort pinned high for the work where the answer matters, and let everything else start wherever it wants. That also means the expensive setting is scoped to the place I chose it, instead of applying to "what is the flag for tar again".

    12
    Share
    Reply

    Answering anonymously — a moderator will review it first.

    Report
  • @back_of_envelope · 2h ago · 2 replies

    They removed the default model setting. The whole thing is a router now and the reason is compute cost, everything that looks cheap gets shunted to Instant no matter what you pick. There is no setting to fix it because there is no setting any more.

    8
    Share
    Reply

    Answering anonymously — a moderator will review it first.

    Report
    • @openkey_owen · 6h ago

      That is not right and it is worth being precise, because this is the version that ends up in a blog post by Friday.

      What did change: the naming. The number is the generation, and Sol / Terra / Luna are capability tiers that move on their own schedule, so "5.6" on its own does not identify a model. Terra and Luna are not selectable in ordinary ChatGPT conversations, which is a real and confusing restriction.

      What did not change: the effort control still exists and choosing Medium or higher still gets you Sol on a paid plan. "I cannot select Terra" and "the default setting was deleted" are two different claims and only the first one holds up. If your settings screen genuinely has no default model entry, screenshot it, because that would be a bug rather than the documented behaviour.

      26
      Share
      Reply

      Answering anonymously — a moderator will review it first.

      Report
  • @flakey_test · 2h ago

    One thing worth checking before you file anything: does it happen on chats started from the home screen and from inside a project, or only one of them? Mine behaved differently depending on where the thread was started, which is exactly the sort of detail that makes a bug report look like nonsense when someone else cannot reproduce it.

    6
    Share
    Reply

    Answering anonymously — a moderator will review it first.

    Report