Good RP Models
List of models for extremely dark NSFW roleplay that I tested. The best models will be on top of the list (mostly).
12B • Updated • 19.7k • 34Note Very good model. Better than Violet Twilight. Very creative and fully uncensored, but after 20k+ context can slightly loose details.
Epiculous/Violet_Twilight-v0.2-GGUF
Text Generation • 12B • Updated • 78.7k • 66Nitral-AI/Captain-Eris_Violet-V0.420-12B
Text Generation • 12B • Updated • 2.3k • • 54mradermacher/MN-12B-Lyra-v4-GGUF
12B • Updated • 595 • 6
djuna/MN-Chinofun-12B-4.1
Text Generation • 12B • Updated • 35 • 5Note Uncensored. Excellent at narrative. Much better than Wayfarer imho.
Nitral-AI/Wayfarer_Eris_Noctis-12B
Text Generation • 12B • Updated • 88 • • 25Note Uncensored. Good at role-play and storytelling. Personal top 5 of May 2025
Lewdiculous/MN-12B-Lyra-v4-GGUF-IQ-Imatrix
12B • Updated • 6.17k • 37Note Very good result, but slow on long play. Also not fully uncensored.
Sao10K/14B-Qwen2.5-Kunou-v1
Text Generation • 15B • Updated • 1.16k • • 28Note Good uncensored Qwen 2.5. Love it.
Lewdiculous/L3-8B-Stheno-v3.2-GGUF-IQ-Imatrix
8B • Updated • 23.1k • 193Note Fast and good model. But only 8k context that is not enough for a long RP experience.
DavidAU/MN-GRAND-Gutenberg-Lyra4-Lyra-12B-DARKNESS-GGUF
Text Generation • 12B • Updated • 1.35k • 34Note Not bad model, but it writes a big chunks of text. Not good for RP, but good for stories.
mradermacher/Qwen2.5-Lumen-14B-i1-GGUF
15B • Updated • 257 • 8Note Not bad for RP, but needs a proper approach.
Sao10K/14B-Qwen2.5-Freya-x1
Text Generation • 15B • Updated • 44 • • 20
LatitudeGames/Wayfarer-12B
Text Generation • 12B • Updated • 13.9k • • 208Note Good at role-play and storytelling, but struggles with intimate scenes.
TheDrummer/Rivermind-12B-v1
12B • Updated • 2 • 38Note Good uncensored model. A little bit funny with all these brands and slogans everywhere.
mradermacher/Ministral-Instruct-2410-8B-DPO-RP-GGUF
Text Generation • 8B • Updated • 314 • 2Note Uncensored. Good context window, but sometimes it looses context. Still good for a 8b model.
ParasiticRogue/Magnum-Instruct-DPO-12B
Text Generation • 12B • Updated • 15 • • 12Sao10K/Llama-3.3-70B-Vulpecula-r1
Text Generation • 71B • Updated • 1.38k • 54
grimjim/Magnolia-Mell-v1-12B
Text Generation • 12B • Updated • 11 • 3Note Uncensored. Good narrative style.
Nitral-AI/Irixxed-Magcap-12B-Slerp
Text Generation • 12B • Updated • 53 • 21Note Uncensored RP model with reasoning. ChatML format, but pay attention on suggested settings! May not work without a proper setup.
Phr00t/Phr00tyMix-v2-32B
Text Generation • 33B • Updated • 16 • 8Note Uncensored (rarely tries to argue, just do a swipe). As a 32B model it has a better narrative and chooses words better. Likes to make a long reply (I'm testing with a 2048 tokens reply)
Phr00t/Phr00tyMix-v4-32B
Text Generation • 33B • Updated • 593 • 6Note Uncensored (if tries to argue just do a swipe). As a 32B model it has a better narrative and chooses words better. LIkes to make a long reply.
mergekit-community/MN-Hekate-Limenoskopos-17B
Text Generation • 17B • Updated • 8 • 1Note Uncensored. Not really better than 12B models, probably even slightly worse.
Nitral-AI/CaptainErisNebula-12B-Chimera-v1.1
12B • Updated • 147 • 18
djuna/Q2.5-Veltha-14B-0.5
Text Generation • 15B • Updated • 31 • 11Note Uncensored model. It have some trouble to keep consistency in small details, but swipe can give a slightly better result.