Model Option Settings
Option Configuration
- You can specify the LLM model and its options in a single line. The format is
model_name [--option1=value] [--option2=value] .... If options are not specified, default values are used.
AI Model Specification Options
| Option | Description | Default |
|---|---|---|
baseurl | Base URL for the API endpoint. Change this when using providers other than OpenRouter. | https://openrouter.ai/api/v1 |
reasoning | Specify reasoning mode for supported models. true, false, none, minimal, low, medium, high, xhigh | (none) |
reasoning_effort | Specify reasoning mode for the reasoning_effort type specification method. | (none) |
kwargs_reasoning | Specify reasoning mode for the chat_template_kwargs.enable_thinking type specification method. | (none) |
temperature | Controls randomness of generation. Lower values make output more deterministic. | (model default) |
top_p | Nucleus sampling parameter. | (model default) |
top_k | Top-k sampling parameter. | (model default) |
frequency_penalty | Penalizes new tokens based on their existing frequency. | (model default) |
presence_penalty | Penalizes new tokens based on whether they appear in the text so far. | (model default) |
repetition_penalty | Penalizes repetition of token sequences. | (model default) |
min_p | Sets the minimum probability for tokens to be considered. | (model default) |
top_a | Filters tokens based on the cumulative probability of the most likely tokens. | (model default) |
max_tokens | Maximum number of tokens to generate. | 10240 |
timeout | Timeout for story generation (milliseconds). | 600000 |
only | (OpenRouter only) Use only a specific provider. | (none) |
strict | For structured output, specify json_schema when true, json_object when false. | false |
stream | Specify whether to use streaming generation. | true |
Examples
- To use DeepSeek V4 Flash 0731:
deepseek/deepseek-v4-flash-0731 - To connect to a local LM Studio instance:
dummy --baseurl=http://127.0.0.1:1234/v1- Even if you don't specify a model name, please include dummy text.
- To use Google AI Studio's Gemini 3.5 Flash lite preview via OpenRouter:
google/gemini-3.5-flash-lite-preview --only=google-ai-studio- To use Google AI Studio directly as a provider:
gemini-3.5-flash-lite --baseurl=https://generativelanguage.googleapis.com/v1beta/openai/
- To use Google AI Studio directly as a provider:
Option Validity
- The options accepted vary by model. Options not accepted by a model will be ignored.
Using Presets
- Instead of specifying the options above, you can configure and use OpenRouter presets, which will be automatically applied during API calls.
Connecting to Non-OpenRouter Providers
- This application uses the OpenAI SDK. Please specify an OpenAI SDK-compatible URL for the BaseURL.
- Connection to all providers claiming compatibility is not guaranteed.