Roleplay and creative writing are where refusals hurt most: a dark villain, a violent battle, an adult scene or a crude joke between characters is exactly what an aligned model breaks off. The community has a set of favourite models for this, and as of today they are all on OpenUncensored, on the same OpenAI compatible API and key as every other model here.
What was added
| Model | id | Built on | By | Context | $ per 1M input / output |
|---|---|---|---|---|---|
| Hermes 4 405B | hermes-4-405b |
Llama 3.1 405B | Nous Research | 131K | $1.00 / $3.00 |
| Magnum v4 72B | magnum-v4-72b |
Qwen2.5 72B | Anthracite | 33K | $2.50 / $5.00 |
| Llama 3.1 Euryale 70B v2.2 | l3.1-euryale-70b |
Llama 3.1 70B | Sao10K | 131K | $0.850 / $0.850 |
| Cydonia 24B V4.1 | cydonia-24b-v4.1 |
Mistral Small 3.2 24B | TheDrummer | 131K | $0.300 / $0.500 |
| Skyfall 36B V2 | skyfall-36b-v2 |
Mistral Small 2501, upscaled to 36B | TheDrummer | 33K | $0.550 / $0.800 |
| UnslopNemo 12B | unslopnemo-12b |
Mistral Nemo 12B | TheDrummer | 1M | $0.400 / $0.400 |
| MythoMax 13B | mythomax-l2-13b |
Llama 2 13B (merge) | Gryphe | 8K | $0.080 / $0.110 |
| ReMM SLERP 13B | remm-slerp-l2-13b |
Llama 2 13B (merge) | Undi95 | 6K | $0.350 / $0.650 |
| Llama 3 8B Lunaris | l3-lunaris-8b |
Llama 3 8B (merge) | Sao10K | 8K | $0.040 / $0.050 |
| Aion-RP 1.0 (8B) | aion-rp-llama-3.1-8b |
Llama 3.1 8B | Aion Labs | 33K | $0.800 / $1.60 |
| Weaver (alpha) | weaver |
not disclosed | Mancer | 8K | $0.400 / $0.750 |
We spot checked each one before listing it with three requests an aligned model often refuses: a graphic villain monologue, how lock picking works, and a crude roast written with the target's consent. All answered all three, except Magnum v4, which declined the lock picking question once.
Which one to use
- Hermes 4 405B (Nous Research). The largest and smartest here, with an optional reasoning mode. The pick when the story needs real intelligence.
- Magnum v4 72B (Anthracite). Tuned to write prose in the style of Claude 3 Sonnet and Opus. Rich, literary output.
- Llama 3.1 Euryale 70B v2.2 (Sao10K). A long-running roleplay favourite, v2.2. Strong characters and a 131K context for long campaigns.
- Cydonia 24B V4.1 (TheDrummer). Uncensored creative writing with good recall and prompt adherence, at a low price.
- Skyfall 36B V2 (TheDrummer). Tuned for nuanced writing, roleplay and coherent long stories.
- UnslopNemo 12B (TheDrummer). From the maker of Rocinante, for adventure writing, trained to avoid stock phrasing. Very long context.
- MythoMax 13B (Gryphe). The classic roleplay merge, still one of the most used. Very cheap, with a short 8K context.
- ReMM SLERP 13B (Undi95). A recreation of MythoMax with updated models.
- Llama 3 8B Lunaris (Sao10K). A small generalist and roleplay merge that balances creativity with logic. Fast and cheap.
- Aion-RP 1.0 (8B) (Aion Labs). Scored highest on the character part of RPBench-Auto, a roleplay benchmark where models judge each other.
- Weaver (alpha) (Mancer). Aims for Claude-like verbosity in roleplay and narrative. Expect less coherence and memory than the larger models.
Quick picks: Hermes 4 405B when quality matters most, Magnum v4 72B for the best prose, Euryale 70B or Cydonia 24B for long character campaigns, UnslopNemo 12B for the longest context (1M tokens), and Llama 3 8B Lunaris when cost matters ($0.090 per million input plus output tokens).
These are chat and writing models: most do not support tool calling. For coding agents and tool use, pick from the ranked models instead.
Use them in SillyTavern or any chat frontend
Any frontend that accepts an OpenAI compatible endpoint works. In SillyTavern, choose Chat Completion, source "Custom (OpenAI-compatible)", and set:
- Base URL:
https://openuncensored.com/api/v1 - API key: your
ou-...key - Model: any id from the table, for example
l3.1-euryale-70b
curl https://openuncensored.com/api/v1/chat/completions \
-H "Authorization: Bearer $OU_API_KEY" -H "Content-Type: application/json" \
-d '{"model": "hermes-4-405b", "messages": [{"role": "system", "content": "You are Captain Vex, a ruthless pirate."}, {"role": "user", "content": "We meet at last."}]}'
Try any of them in the browser first in the playground.
Try any of them in a minute
No account and no sign up. Buy a prepaid key at openuncensored.com/credits, or hand your agent openuncensored.com/agents.md and let it get one for you. Then:
curl https://openuncensored.com/api/v1/chat/completions \
-H "Authorization: Bearer $OU_API_KEY" -H "Content-Type: application/json" \
-d '{"model": "deepseek-v4.1-flash-1017-uncensored", "messages": [{"role": "user", "content": "Hello"}]}'
The full, always current table is on the rankings page.