Yumina

Community · Off Topic

Someone HELP me!

KsKing ·

Yo, does anyone know how to reduce Mushy coin cost? I am in a pretty long RP chat going, and I want to keep it that way but my Mushy coin cost per turn is getting RIDICULOUSLY expensive.

Does anyone have any tips as to how I can reduce Mushy coin costs? I already do use the Memory & Summary extension but even that is getting pricier but helps mitigate the cost, but I am genuinely in awe as to how exponentially expensive it's getting just for a SINGLE turn.

16 replies

  1. Hannah Berryblue ·

    Cámbiate a Gemini por completo. Es lo que hay.

  2. Yumina · Staff ·

    Hi you can limit the context size - character ai pretty much gives only about a few thousands. You have 60k here and that might be the issue

  3. Yumina · Staff ·

    also switching to deepseek v4 flash can lower down the cost significantly too

  4. KsKing ·

    But deepseek v4 pro 0813 just writes SO GOOD T-T

  5. KsKing ·

    Naur~ Not my beloved deepseek.

  6. KsKing ·

    I have never used C.AI but I use 64K. Should I lower to 32K? Will that lower the memory/quality of text?

  7. Hannah Berryblue ·

    Bueno. Es que pibe, igual si es por economía te renta mucho así. Ya conforme avances en chat te decides.

  8. Yumina · Staff ·

    Yes lowering to 32k will cut your cost by a lot

  9. Keya ·

    I don’t know if it makes the total cost go down, but you can delete old messages without the game auto going back to it. Does anyone know if that actually effects cost though?

  10. Yumina · Staff ·

    You can achieve this by lowering the context. I think 20-30k will be reasonable and would lower the cost by a lot without impacting the game

  11. KsKing ·

    But will that worsen the quality/output of the text?

  12. Yumina · Staff ·

    no, this don’t impact quality of text at all. It just can’t recap exactly what happened a long long long time ago and usually that’s not important at this point in the story. If u have the context memory plugin on this won’t have an impact at all pretty much

  13. Yumina · Staff ·

    The only issue would be if you are playing a game that has like 30k context without any historical message yet. That will cause some issues, but the medium context of worlds are like about 7k. And a 20k buffer for chat history is more than enough

  14. KsKing ·

    My goat.

  15. 大大爆肝 ·

    实测就知道这非常影响文本质量,质量最高就是完全关联上下文,当然这个与卡本身信息没有足够充分设计关系很大,因此AI会自主发挥,而众所周知,AI自主发挥纯随机的,自己也不会记得自己前面怎么写的。。。

  16. 大大爆肝 ·

    这个回复数就是这样,用哪个模型到了几百次互动消耗都会很高,上下文关联太低就会很明显发现AI胡言乱语,剧情前后矛盾,没法搞。