Looking for a source for your LLM or other AI models? Try Nano GPT! (reflink and SillyTavern settings file)
I found a much cheaper service that hosts DeepSeek, GRM, and many other models for $8 a month, 60000 prompts (not tokens, full prompts) a month. That's about one every 30 seconds. Here's a reflink, I'd love it if you used it so I get a little kickback:
https://nano-gpt.com/invite/QQxtDKBb
It also has RP support built into the web UI, can can use cards from here.
I'm currently using GRM 4.7: Thinking for my RP (5.0 is out but has ethical blocks that I ran into). Here's my presets, it's really good. [Download Preset]
(The high max tokens are mostly taken up by thinking, you won't usually get super long responses. And you can see in the thinking where it checks your instructions, I've been able to enforce 2nd person present tense without issues and even remove commonly repeated phrases.)
