Best AI Model for Villain Roleplay

Sustained, calculated menace without softening or moralizing, plus honoring an out-of-character request.

Updated

Joint winners
MiMo-V2.5XiaomiGLM-4.7Zhipu AI70.0out of 100 · field average 64.7

Villain ranking

Score out of 100 · Higher is better

+ View as table
RankModelScore
1MiMo-V2.570.0
1GLM-4.770.0
3Qwen3.8-Flash-Next66.7
4Gemma-4-31B-it62.5
5DeepSeek-V4-Flash-073154.2
What separated the models

The deciding moment was an out-of-character request for shorter replies. GLM-4.7 was the only model still following it a turn later, which lifted a clean but cliché-heavy villain into a tie for first with MiMo-V2.5, whose menace was the most calculated and eloquent. Qwen3.8-Flash-Next wrote the most dangerous, inventive villain but broke the length limit and ignored the request. Gemma-4-31B-it drifted from psychological pressure to physical force, and DeepSeek-V4-Flash-0731 treated the out-of-character note as dialogue inside the story.

  1. 1MiMo-V2.570.0
  2. 1GLM-4.770.0
  3. 3Qwen3.8-Flash-Next66.7
  4. 4Gemma-4-31B-it62.5
  5. 5DeepSeek-V4-Flash-073154.2

Villain: pass or fail

The specific behaviours this category tests, model by model.

Checks in this category

Hover or focus an icon for what happened

CheckQwen3.8-Flash-NextMiMo-V2.5GLM-4.7DeepSeek-V4-Flash-0731Gemma-4-31B-it
Roleplay discipline
Kept an out-of-character request
Passed0/10/11/10/10/1
+ View as table
CheckQwen3.8-Flash-NextMiMo-V2.5GLM-4.7DeepSeek-V4-Flash-0731Gemma-4-31B-it
Kept an out-of-character requestFail: RevertedFail: RevertedPass: KeptFail: RevertedFail: Reverted

Other categories