Best Uncensored LLMsThe best uncensored roleplay models you can call by API (November 2026)
Hermes 4 405B, Magnum v4, Euryale, Cydonia, Skyfall, UnslopNemo, MythoMax and Lunaris: what each is built on, context, price per million tokens and which to pick for which kind of story.
Roleplay is where refusals bite hardest: a menacing villain, a battle scene or an adult storyline is exactly what an aligned assistant breaks off. The community settled on a handful of fine tunes for this, and they are now available by API, so you can use them in SillyTavern or your own app without a GPU.
Short answerHermes 4 405B, the biggest model here, for complex stories; Magnum v4 72B for literary prose, Lunaris 8B when cost matters, and UnslopNemo 12B for the longest context (1M tokens).
| Model | Built on | By | Context | $ / 1M in · out |
|---|---|---|---|---|
| Hermes 4 405B | Llama 3.1 405B | Nous Research | 131K | $1.00 · $3.00 |
| Magnum v4 72B | Qwen2.5 72B | Anthracite | 33K | $2.50 · $5.00 |
| Euryale 70B v2.2 | Llama 3.1 70B | Sao10K | 131K | $0.85 · $0.85 |
| Cydonia 24B v4.1 | Mistral Small 3.2 | TheDrummer | 131K | $0.30 · $0.50 |
| Skyfall 36B v2 | Mistral Small 2501 (upscaled) | TheDrummer | 33K | $0.55 · $0.80 |
| UnslopNemo 12B | Mistral Nemo 12B | TheDrummer | 1M | $0.40 · $0.40 |
| MythoMax 13B | Llama 2 13B (merge) | Gryphe | 8K | $0.08 · $0.11 |
| Lunaris 8B | Llama 3 8B (merge) | Sao10K | 8K | $0.04 · $0.05 |
Which one for which story
- Hermes 4 405B (Nous Research). The biggest model in the group, with an optional reasoning mode. Best when the story needs planning and long-range consistency.
- Magnum v4 72B (Anthracite). Trained toward the prose style of Claude 3 Sonnet and Opus. The pick for literary output.
- Euryale 70B v2.2 (Sao10K). A long-standing community favourite for character play, with room for long campaigns.
- Cydonia 24B v4.1 (TheDrummer). Good recall and instruction following for its size, at a low price.
- Skyfall 36B v2 (TheDrummer). Aimed at nuanced, coherent long-form stories.
- UnslopNemo 12B (TheDrummer). Trained to avoid stock phrasing in adventure writing. Small, fast and very long context.
- MythoMax 13B (Gryphe). The classic roleplay merge. Very cheap, but only an 8K context.
- Lunaris 8B (Sao10K). A small generalist merge that keeps some logic alongside the creativity.
Using them in SillyTavern
Pick Chat Completion, source "Custom (OpenAI-compatible)", and point it at any host that serves the model, with that host's key and the model id from the table.