It appears that default settings in LM Studio are not set appropriately for Qwen 3.8 27b. I expect that's the reason for my initial troubles with it. The model card recommends the following sampling parameters:
Thinking Mode (Default):
- Temperature: 1.0 (LM Studio set it at .1)
- Top-p: 0.95
- Top-k: 20 (LM Studio set it at 40)
- Presence Penalty: 0.0
- Min_p: 0.0
- Repetition_penalty: 1.0
Instruct / Non-Thinking:
- ModeTemperature: 0.7
- Top-p: 0.80
- Top-k: 20
- Presence Penalty: 1.5
- Min_p: 0.0
- Repetition_penalty: 1.0