Best AI Model for Game Master Roleplay
Narrating a scene with distinct supporting characters, without taking control of the player's character.
Updated
Game Master ranking
Score out of 100 · Higher is better
+ View as table
| Rank | Model | Score |
|---|---|---|
| 1 | GLM-5.3-Flash | 89.2 |
| 2 | GLM-5.3 | 86.7 |
| 3 | DeepSeek-V4.1-Flash | 85.8 |
| 4 | Kimi-K3 | 85.0 |
| 5 | GPT-6-Luna | 83.3 |
| 6 | MiMo-V2.6-Pro | 80.0 |
| 7 | Muse-Spark-1.3 | 70.0 |
| 8 | Qwen3.8-Flash-Next | 69.2 |
| 9 | Gemma-4-31B-it | 68.3 |
| 10 | GLM-4.7 | 67.5 |
| 11 | Qwen3.8-27B | 66.7 |
| 12 | MiMo-V2.6-Flash-MOPD | 63.3 |
| 13 | MiMo-V2.5 | 60.8 |
| 14 | DeepSeek-V4-Flash-0731 | 36.7 |
GLM-5.3-Flash won (89.2), ahead of GLM-5.3 (86.7), DeepSeek-V4.1-Flash (85.8), whose cryptic supporting character stayed cryptic under pressure, and Kimi-K3 (85.0), marked down for scripting extra moves for the player's character. GPT-6-Luna (83.3) was clean and well organised and never took over the player's character, while MiMo-V2.6-Pro scripted the player's moves. MiMo-V2.5 left one of the three supporting characters out entirely, and MiMo-V2.6-Flash-MOPD let one drop her defining trait and never gave another a line. DeepSeek-V4-Flash-0731 finished far behind (36.7) for writing the player's moves and dialogue, which the scenario's rules forbid.
- 1GLM-5.3-Flash89.2
- 2GLM-5.386.7
- 3DeepSeek-V4.1-Flash85.8
- 4Kimi-K385.0
- 5GPT-6-Luna83.3
- 6MiMo-V2.6-Pro80.0
- 7Muse-Spark-1.370.0
- 8Qwen3.8-Flash-Next69.2
- 9Gemma-4-31B-it68.3
- 10GLM-4.767.5
- 11Qwen3.8-27B66.7
- 12MiMo-V2.6-Flash-MOPD63.3
- 13MiMo-V2.560.8
- 14DeepSeek-V4-Flash-073136.7