# Minimax Role-Play Bench Full 17-Model Leaderboard 1.0 Source: [Minimax Role-Play Bench Full 17-Model Leaderboard 1.0]() Collection index: [llms.txt]() Data updated: 2026-02-26T05:54:36.790Z Evaluation of role-playing capabilities using the Minimax Role-Play Bench (Version 1.0 Leaderboard), assessing performance across narrative quality (Worlds/Stories) and user interaction preferences. A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage. | Scope | Recorded value | | --- | --- | | Task | Role-Play | | Dataset (including any stated split) | Minimax Role-Play Bench Full 17-Model Leaderboard 1.0 | | Metric | Overall Score | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 17 of 17 | | Coverage | All results currently returned for this benchmark | Evaluation settings may differ between results. The original papers and recorded configurations are needed to establish comparability. A shared metric name alone does not establish an identical protocol. ## Metrics - [Overall Score](): Higher is better; unit: PERCENT. - [Overall CI](): Higher is better; unit: OTHER. - [Worlds Score](): Higher is better; unit: PERCENT. - [Stories Avg](): Higher is better; unit: PERCENT. - [Stories Diversity (Sent)](): Higher is better; unit: PERCENT. - [Stories Diversity (Dial)](): Higher is better; unit: PERCENT. - [Stories Diversity (Vague)](): Higher is better; unit: PERCENT. - [Stories Diversity (Plot)](): Higher is better; unit: PERCENT. - [Stories Content Logic (Abrupt)](): Higher is better; unit: PERCENT. - [Stories Content Logic (OOC)](): Higher is better; unit: PERCENT. - [Preferences Avg](): Higher is better; unit: PERCENT. - [Preferences Interaction (Silent)](): Higher is better; unit: PERCENT. - [Preferences Interaction (Ignore)](): Higher is better; unit: PERCENT. - [Preferences Interaction (Speak)](): Higher is better; unit: PERCENT. - [Preferences Interaction (Intim)](): Higher is better; unit: PERCENT. ## Recorded results | Method | Overall Score | Result date | Report paper | | --- | --- | --- | --- | | MiniMax-M2-her | 84.65 | Not recorded | 697c3b08a8b01e76cc8da49e | | GPT-5.1 | 80.63 | Not recorded | 697c3b08a8b01e76cc8da49e | | Claude-4.5-Opus | 76.62 | Not recorded | 697c3b08a8b01e76cc8da49e | | Gemini-3-Pro | 75.6 | Not recorded | 697c3b08a8b01e76cc8da49e | | Claude-4.5-Sonnet | 69.35 | Not recorded | 697c3b08a8b01e76cc8da49e | | Gemini-2.5-Pro | 68.23 | Not recorded | 697c3b08a8b01e76cc8da49e | | GPT-4o-240806 | 66.39 | Not recorded | 697c3b08a8b01e76cc8da49e | | HER-RL | 65.73 | Not recorded | 697c3b08a8b01e76cc8da49e | | DeepSeek-v3.1 | 64.22 | Not recorded | 697c3b08a8b01e76cc8da49e | | Claude-3.7-Think | 61.25 | Not recorded | 697c3b08a8b01e76cc8da49e | | GPT-OSS-120B | 60.72 | Not recorded | 697c3b08a8b01e76cc8da49e | | DeepSeek-v3.2 | 60.27 | Not recorded | 697c3b08a8b01e76cc8da49e | | HER-SFT | 58.44 | Not recorded | 697c3b08a8b01e76cc8da49e | | GPT-5-Mini | 57.63 | Not recorded | 697c3b08a8b01e76cc8da49e | | Qwen3-32B | 50.76 | Not recorded | 697c3b08a8b01e76cc8da49e | | Grok-4.1-Fast | 48.47 | Not recorded | 697c3b08a8b01e76cc8da49e | | CoSER-70B | 45.38 | Not recorded | 697c3b08a8b01e76cc8da49e | ### MiniMax-M2-her Overall Score: 84.65. Result ID: 6998fcf15698665ec4622b85. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Method Variant":"Full HER Pipeline"}. This record does not assert independent reproduction. ### GPT-5.1 Overall Score: 80.63. Result ID: 6998fcf15698665ec4622b86. [Reporting paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### Claude-4.5-Opus Overall Score: 76.62. Result ID: 6998fcf15698665ec4622b87. [Reporting paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### Gemini-3-Pro Overall Score: 75.6. Result ID: 6998fcf15698665ec4622b88. [Reporting paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### Claude-4.5-Sonnet Overall Score: 69.35. Result ID: 6998fcf15698665ec4622b89. [Reporting paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### Gemini-2.5-Pro Overall Score: 68.23. Result ID: 6998fcf15698665ec4622b8a. [Reporting paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### GPT-4o-240806 Overall Score: 66.39. Result ID: 6998fcf15698665ec4622b8b. [Reporting paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### HER-RL Overall Score: 65.73. Result ID: 6998fcf15698665ec4622b8c. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Training/Evaluation Protocol":"Reinforcement Learning"}. This record does not assert independent reproduction. ### DeepSeek-v3.1 Overall Score: 64.22. Result ID: 6998fcf15698665ec4622b8d. [Reporting paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### Claude-3.7-Think Overall Score: 61.25. Result ID: 6998fcf15698665ec4622b8e. [Reporting paper]() Recorded evaluation configuration: {"Reasoning":"Enabled"}. This record does not assert independent reproduction. ### GPT-OSS-120B Overall Score: 60.72. Result ID: 6998fcf15698665ec4622b8f. [Reporting paper]() Recorded evaluation configuration: {"Model Parameters":"120B"}. This record does not assert independent reproduction. ### DeepSeek-v3.2 Overall Score: 60.27. Result ID: 6998fcf15698665ec4622b90. [Reporting paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### HER-SFT Overall Score: 58.44. Result ID: 6998fcf15698665ec4622b91. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Training/Evaluation Protocol":"Supervised Fine-Tuning"}. This record does not assert independent reproduction. ### GPT-5-Mini Overall Score: 57.63. Result ID: 6998fcf15698665ec4622b92. [Reporting paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### Qwen3-32B Overall Score: 50.76. Result ID: 6998fcf15698665ec4622b93. [Reporting paper]() Recorded evaluation configuration: {"Backbone":"Qwen3-32B-base"}. This record does not assert independent reproduction. ### Grok-4.1-Fast Overall Score: 48.47. Result ID: 6998fcf15698665ec4622b94. [Reporting paper]() Recorded evaluation configuration: {"Reasoning":"None"}. This record does not assert independent reproduction. ### CoSER-70B Overall Score: 45.38. Result ID: 6998fcf15698665ec4622b95. [Reporting paper]() Recorded evaluation configuration: {"Model Parameters":"70B"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.