runjake 22 hours ago

Sonnet 5 Medium and GPT 5.5 Medium for coding.

I drop down to a Mini model for file-based tasks.

Codex keeps trying to revert me back to 5.6 but 5.6 seems to use too many tokens to do the same or slightly inferior job.

I am only on multiple $20 plans.

  • neverenderr 22 hours ago

    I see. I am just using the $20 claude paln, so I can't use GPT.

igor_nast 1 day ago

Opus - planning/review + gpt spark med - implementation. Works better at lower cost.

  • neverenderr 1 day ago

    I don't have a subscription to Codex, but it's good to know.

leros 23 hours ago

I've been using Fable for planning and Opus for coding

  • neverenderr 23 hours ago

    What plan are you on? Max?

    • leros 20 hours ago

      Max 5X. I have plenty of headroom coding using Opus all day so I figured I may as well use Fable a bit. I have found Fable to do a much better job of understanding a codebase and avoiding subtle bugs.

      • neverenderr 19 hours ago

        Oh man, I haven't used fable that much, so I can't say much, but based on your experience, I think it seems pretty good. Also, I am on the pro plan, so I can't afford to use fable xD.

        • leros 19 hours ago

          Opus has been great for me on building simple features or adding new things. Fable has been significantly better for things like "there are subtle bugs in this complex interface I can't figure out, find them". Opus gets a rough idea for some. Fable actually finds all the tiny subtle issues.

          • neverenderr 18 hours ago

            Opus is good for planning, and fable seems interesting now.