ComfyUI Node
LM Studio Options
A ComfyUI node in LM Studio with 10 inputs and 1 output.
LM Studio Options
- options
◄temperature0.20►
◄max_tokens512►
◄seed0►
◄top_p1.00►
◄top_k0►
◄frequency_penalty0.00►
◄presence_penalty0.00►
◄repeat_penalty1.00►
◄strip_thinking_tagsfalse►
◄stop►
CategoryLM Studio
Inputs (10)
| Name | Type | Default | Description |
|---|---|---|---|
| temperature | FLOAT | 0.200–2 | Sampling temperature. Higher = more random; lower = more deterministic. |
| max_tokens | INT | 512-1–32768 | Maximum number of new tokens to generate for the response. Use -1 for unlimited (recommended for thinking models). |
| seed | INT | 00–2147483647 | Optional random seed for reproducible sampling. Use 0 to disable and let the model choose. |
| top_p | FLOAT | 1.000–1 | Nucleus sampling: consider tokens with cumulative probability up to top_p. Set to 1.0 to disable. |
| top_k | INT | 00–2048 | Top-K sampling: only consider the top_k most likely tokens. Set to 0 to disable. |
| frequency_penalty | FLOAT | 0.000–2 | Penalize tokens proportionally to how often they have appeared. Helps reduce repetition. |
| presence_penalty | FLOAT | 0.000–2 | Penalize tokens if they have appeared at all. Encourages introducing new topics. |
| repeat_penalty | FLOAT | 1.000–2 | Generic repetition penalty (model/engine specific). 1.0 means no penalty. |
| strip_thinking_tags | BOOLEAN | false | Remove <thinking>...</thinking> tags and their content from the output. Useful for getting clean captions without reasoning text. |
| stopopt | STRING | Comma-separated stop strings. Generation will stop when any is encountered. Leave blank for none. |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| options | LMSTUDIO_OPTIONS | — |