Presets

SillyTavern presets and settings.. Tested configurations for a variety of models to improve your AI roleplay experience.

Model
All 7 Claude Opus 4.6 1 DeepSeek V3.2 1 DeepSeek V4 Flash 1 Gemini 3.5 Flash 1 GLM 5.2 1 GLM-4.7 1 GLM-5 1

GLM 5.2 Presets for SillyTavern

~ Hutch ~ Leave a comment

GLM 5.2 Presets for SillyTavern

On a day such as this, one could not have expected better news for the GLM faithful. GLM 5.2 just dropped on OpenRouter the second week of June, and the early community response has been strong. GLM 5.1 was already ranked #2 overall by the Freaky Frankenstein creator (per their tier list, output 9/10, censorship: zero), and 5.2 is just more of that on more headroom. The chain-of-thought architecture is the real story though. Give it a tiered preset and it processes each layer before it spits out a response, like a DM checking session notes. That combination is what makes GLM feel less like a model and more like a partner.

Continue reading →

DeepSeek V4 Flash Presets for SillyTavern

~ Hutch ~ Leave a comment

DeepSeek V4 Flash Presets for SillyTavern

DeepSeek V4 Flash had a rocky launch, and I am not going to sugarcoat that. Community reports cited POV confusion, stereotypical RP output, and weak prompt adherence. The big wrinkle is that V4 ships with built-in RP instructions that actively fight custom presets, which is why it needs special handling. The good news: V4 Flash is “surprisingly good” per the community as of June 2026, and DeepSeek has committed to ongoing roleplay-specific tuning based on community feedback. The 1M context window and permanent 75% price reduction make it one of the most accessible frontier models for marathon sessions. Use V3.2 when you need reliability, V4 Flash when you need 1M context or cost.

Continue reading →

DeepSeek V3.2 Presets for SillyTavern

~ Hutch ~ Leave a comment

DeepSeek V3.2 Presets for SillyTavern

I tried V3.2 again last month after bouncing off it earlier in the year, and I am not going to lie, it caught me off guard. “Apparently all this model needed was a quality preset. My goodness can it cook” is the vibe in the megathreads right now, and I’m inclined to agree. The community has it ranked #3 overall (per the Freaky Frankenstein tier list as of June 2026), the FF creator uses it interchangeably with GLM 5.1 depending on scene type, and the prose is vivid in a way the front-page OpenRouter rankings don’t really capture. The one real limitation is the 163K context window, which is lower than the 1M-token models, so very long sessions can fall off. That’s a tradeoff, not a dealbreaker.

Continue reading →

Claude Opus 4.6 Presets for SillyTavern

~ Hutch ~ Leave a comment

Claude Opus 4.6 Presets for SillyTavern

Look. I use Claude (sparingly). I have used it for a long time (sparingly!), and I am not going to sit here and pretend the bill doesn’t hurt. The prose is still the best in the business, the emotional range is something the other models can’t touch, and the way it handles character dynamics without you having to spell them out is a genuine superpower. The friction, beyond the price, is that Anthropic has been quietly adding new API-level safeguards to older Claude models, so things that worked last month may suddenly refuse. That part is annoying. It is what it is.

Continue reading →

Gemini 3.5 Flash Presets for SillyTavern

~ Hutch ~ Leave a comment

Gemini 3.5 Flash Presets for SillyTavern

Gemini 3.5 Flash presets are plentiful. If you’ve used Gemini 3.5 Flash for any length of time, you already know the deal. It’s fast, it’s cheap, and the prose is the best the Gemini line has produced to date. “Less ‘Geminisms'” is the consensus, and that tracks with what I’ve seen in long sessions. The main practical tip is to turn off streaming on AI Studio, otherwise the API-level filter will block your message sends mid-stream. On the censorship question, Flash is less restrictive than previous Gemini models for NSFW, but violence has some filter friction on AI Studio specifically. NanoGPT and OpenRouter tend to be cleaner. The 1M context window and Flash-tier pricing make it a strong value pick.

Continue reading →

GLM-5 Presets

~ Hutch ~ 1 Comment

GLM-5 Presets – Stab’s Directive Hierarchy for SillyTavern

Link to Github Here!

Before anything is said, yes: I am a Z.ai simp. If you’re familiar with GLM and used Stab’s Directives on GLM-4.7, this’ll feel like the natural next form. It’s the same greatness you can come to expect from Zorgonatis at this point. That said, it’s pretty much similar, just better in almost every way. With GLM‑5’s extra headroom, there’s better handling of character dynamics, stronger continuity, and less chatter or deity‑worship of the user mid‑scene. The dialogue also flows way better in a way that feels more natural, or how you think a character would respond in a given situation. GLM-5 itself is just tight all around (..That was inappropriate), and Stab’s Directive GLM-5 preset tightens it even further (Enough!). This makes for a fun, well-structured, organized, and consistent RP session. Definitely one of the better GLM-5 Presets I’ve used. Also, it’s good as a Gemini roleplay preset. If you use that.

Continue reading →

GLM-4.7 Presets

~ Hutch ~ 5 Comments

GLM-4.7 – Stab’s Directive Hierarchy for SillyTavern and GLM-4.7 / Gemini 3.0

Link to Github Here!

GLM-5 Presets

Out of all the GLM-4.7 presets I’ve used so far, this is my go-to. I’m a big fan. It’s well-structured, and naturally amplifies the way GLM’s reasoning works; which is based off chain-of-thought (reasoning step-by-step). The preset has several ‘tiered’ rules, that GLM-4.7 naturally follows. The output is more deliberate and consistent (think of it like a DM reviewing notes before a session). The tiers are comprised of:

Continue reading →