TheDrummer/Anubis-70B-v1.1

71B • Updated Jun 29 • 955 • 20

Note Anubus and Lemonades, all great models for Rping and storywriting from what i am seeing. Perform quite nicely in SillyTavern/TavernAI. Handles multiple characters and RP within an RP like a champ! These three types of models are my current goto. It's very easy to forget you aren't replying to a real person.

DavidAU/Qwen2.5-Godzilla-Coder-51B-128k

Text Generation • 51B • Updated 2 days ago • 15 • 1

Note I have NOT ran this through it's paces, but i did ask it some technical questions and have it generate some code that looked more-or-less correct, even included comments; It also did halfway decent in RPing too.

Aurore-Reveil/Austral-Qwen3-235B

235B • Updated Jun 28 • 16 • 2

Note Only tried the Q2 variant for what i can fit in memory, but it did a VERY good job analyzing a complex story getting details most models miss; As well as resuming where it left off keeping track of who, what they were doing, holding, etc. compared to other models. I want to say it does slightly worse vs 70b models in RPing but that's probably the Quanitization, otherwise it's on par with text/storygen, and performs about the same speed.

zerofata/MS3.2-PaintedFantasy-Visage-33B

Text Generation • 34B • Updated 29 days ago • 514 • 17

Note Tried it and was a very nice mid-model that has very nice language and conversation skills. However i saw heavy repeats of things in an RP. I need to test it more but it seems like a fine mid-sized model. (in said RP above the swirling code vortex and warning of networking protocols being detected got a little too annoying for my taste. Also in said RP every time something was brought up new 'encrypted folders' were made to facilitate said neuro interface. It might just be the character card)

spacewars123/Space-Wars-24B-v1.00b

Text Generation • 24B • Updated May 27 • 125 • • 2

Note Very nice, tried not long ago as one of my first larger RPing models. Leans more into sci-fi knowledge/aspects but if memory serves me it did nice overall otherwise and totally sufficient for fantasy.

DavidAU/L3.1-Dark-Planet-SpinFire-Uncensored-8B-GGUF

Text Generation • 8B • Updated 5 days ago • 1.16k • 26

Note A small model for writing/RPing and conversation if you have a weaker machine (weaker as in 16Gb or less memory, or 4Gb/less video card). Struggles with advanced concepts like multiple characters and roleplay within a roleplay. You're also likely to see repeated phrases.

Goekdeniz-Guelmez/Qwen3-0.6B-gabliterated

Text Generation • 0.6B • Updated 15 days ago • 72 • 7

Note Fast and lightweight! Looks promising for a minimal model for even laptops, and for doing a CLI passthrough for processing. Warning: Slows down immensely after 2k token inputs. Before that it's lightning fast and easily fits in a video card.

ockerman0/AnubisLemonade-70B-v1

Text Generation • 71B • Updated about 1 month ago • 16 • 2

Note Anubus and Lemonades, all great models for Rping and storywriting from what i am seeing. Perform quite nicely in SillyTavern/TavernAI. Handles multiple characters and RP within an RP like a champ!

zerofata/L3.3-GeneticLemonade-Unleashed-v3-70B

Text Generation • 71B • Updated Jun 6 • 853 • • 18

Note Anubus and Lemonades, all great models for Rping and storywriting from what i am seeing. Perform quite nicely in SillyTavern/TavernAI. Handles multiple characters and RP within an RP like a champ!

Tarek07/Dungeonmaster-V2.2-Expanded-LLaMa-70B

Text Generation • 71B • Updated Mar 30 • 895 • 13

Note Another nice model!

jukofyork/command-r-35b-writer

Text Generation • 35B • Updated 22 days ago • 103 • 1

Note This model is marked as *'experimental'* but with Temperature between 0.7 i am having a decent time with it's output on SillyTavern. A note, i randomly get Japanese or Arabic words injected as half a word. Rare enough it's not a issue if you rewrite.

OddTheGreat/Mechanism_24B_V.1

Text Generation • 24B • Updated 22 days ago • 42 • 1

Note This model starts good and with short context windows (4k) it does wonderfully. Larger context and it starts dropping common qualifiers and output greatly decreases to half of what it was when it started. Probably Space Wars is a little better, but this one isn't bad at all for a small model.

BruhzWater/Edens-Fall-L3.3-70b-0.3b

Text Generation • 71B • Updated 8 days ago • 12 • 2

Note Promising 70B model. However it felt slower as it went along further in. At about 9k-10k tokens in, it slowed to a crawl. Not sure if that's my fault, or the model, but before that point it performed magnanimously. Tried the c variant as well, it has the same slowdown problem (though a bit slower) and stops somewhere around 10k. If you don't mind say a 4k context window this will do really well.