Skip to main content

There's no understanding of what is "GPT-5.4" and "GPT-5.5"

might be an ez fix.

  • ChatGPT.com serve GPT-5.4 mini, GPT-5.4, GPT-5.5 Instant, GPT-5.5 Thinking, GPT-5.5 Pro. No “GPT-5.4 Pro” btw.

  • openrouter.ai serves GPT-5.4, GPT-5.4 Pro, GPT-5.5, GPT-5.5 Pro.

  • t3.chat serves GPT-5.4 and GPT-5.5, with varied reasoning. The “knowledge cutoff” and “added on” of these 2 models don’t match any of the 4 models on openrouter.

So i don’t understand what’s going on with GPTs and what i’m using.

E.g. if i were to trust openrouter, there’s legit GPT-5.4 Pro and GPT-5.5 Pro. Are they accessible on t3.chat? If so, how?

Please make this clear in the UI. Thank you in advance.

Status: Completed5 comments

Log in to comment and vote

Comments5

  • Mark Florkowski

    Team

    I’ll have a look at the knowledge cutoffs and make sure they are all up to date. (Often at launch this information is not readily available, so we set based upon the last model from the lab)

    GPT-5.4 Pro and GPT-5.5 Pro. Are they accessible on t3.chat?

    5.5 pro is available if you bring your own API Key:

    “Added on” dates are not expected to necessarily line up, as they are the dates the model was added to T3 Chat, not to the upstream providers.

    • Amber Fuelcell

      oh i see. GPT-5.5 Pro isn’t even showing up as an option on the device where i don’t have the key. But where i have the key, it does.

      • Mark Florkowski

        Team

        Yeah, we don’t show the option if you do not have a key set upp to avoid confusion.

        • Amber Fuelcell

          super weird, hiding that exactly caused the confusion for me. Cuz the rest i understood after that. But e.g. it shows where i need a key for higher reasoning for GPT-5.5 and Opus 4.8. Explicit should be less confusing than implicit, no? :)

          • Mark Florkowski

            Team

            Most people do not need the “Pro” models, and having them visible but locked behind an API key would lead to increased support questions for little gain.