AI models & API keys
Connect direct providers or OpenRouter and understand the model settings used by the plugin.
Match your key to the connection
| Selection | Settings field |
|---|---|
| Any text model marked (OpenRouter), and OpenRouter images | OpenRouter API Key |
| Direct OpenAI text models | OpenAI API key |
| Direct Anthropic models | Anthropic Claude API Key |
| Direct Google models | Google Gemini API Key |
| Direct xAI models | xAI API Key |
| Midjourney or Flux through CometAPI | CometAPI API Key |
| Midjourney or Flux through GoAPI | GoAPI API Key |
| Freepik | Freepik API Key |
| Stable Diffusion | Stability AI API key |
Save the keys in Settings, then choose the provider/model in your campaign. A key from one service does not authorize requests to another. Direct and OpenRouter versions can have different access, pricing and output limits.
Click to enlargeText-model mappings in the supplied source
This table documents the non-deprecated AIAB_ENGINES entries in the supplied plugin file. These are plugin mappings and configured token budgets, not a live provider catalog or a guarantee of provider availability. The API may enforce a smaller output limit.
| Plugin selection | Configured model ID | Configured token budget |
|---|---|---|
| OpenAI GPT | gpt-5.5 | 128,000 |
| OpenAI GPT mini | gpt-5.4-mini | 128,000 |
| OpenAI GPT-4o | gpt-4o | 16,384 |
| OpenAI GPT-4o mini | gpt-4o-mini | 16,384 |
| Google Gemini Pro | gemini-3.1-pro-preview | 65,536 |
| Google Gemini Flash | gemini-3.5-flash | 65,536 |
| Anthropic Claude Haiku | claude-haiku-4-5 | 65,536 |
| Anthropic Claude Sonnet | claude-sonnet-4-6 | 65,536 |
| Anthropic Claude Opus | claude-opus-4-7 | 131,072 |
| xAI Grok | grok-4.3 | 262,144 |
| xAI Grok mini | grok-3-mini | 131,072 |
| OpenAI GPT (OpenRouter) | openai/gpt-5.5 | 128,000 |
| OpenAI GPT mini (OpenRouter) | openai/gpt-5.4-mini | 128,000 |
| OpenAI GPT-4o (OpenRouter) | openai/gpt-4o | 16,384 |
| OpenAI GPT-4o mini (OpenRouter) | openai/gpt-4o-mini | 16,384 |
| Google Gemini Pro (OpenRouter) | ~google/gemini-pro-latest | 65,536 |
| Google Gemini Flash (OpenRouter) | ~google/gemini-flash-latest | 65,536 |
| Anthropic Claude Haiku (OpenRouter) | ~anthropic/claude-haiku-latest | 65,536 |
| Anthropic Claude Sonnet (OpenRouter) | ~anthropic/claude-sonnet-latest | 65,536 |
| Anthropic Claude Opus (OpenRouter) | ~anthropic/claude-opus-latest | 131,072 |
| xAI Grok (OpenRouter) | x-ai/grok-4.3 | 1,000,000 |
| DeepSeek (OpenRouter) | deepseek/deepseek-v4-pro | 1,000,000 |
| Mistral Medium (OpenRouter) | mistralai/mistral-medium-3-5 | 262,100 |
Model labels can map to newer versions without changing the campaign label. Legacy entries marked deprecated remain in the code; avoid relying on a retired model for new campaigns. Check the provider account if a model is rejected.
Tokens and practical article size
Max tokens controls the requested output budget for each text section. Outline, title, translation and other supporting requests can use different budgets. Tokens are not words, and context or output limits vary by model. Large budgets can increase cost and runtime.
Live search compatibility
The campaign Live search checkbox applies only to OpenRouter text models. The supplied code appends :online to the selected model ID. It has no effect on direct OpenAI, Anthropic, Google or xAI connections.
OpenRouter currently describes the :online variant as deprecated in its web search documentation. The supplied plugin uses this older mechanism. Test it with your account; if rejected, disable Live search and contact support. Search can incur additional provider charges.
Choose through a small test
Generate one draft in your target language with the intended prompts. Compare output quality, request time and provider usage. Increase section counts or add images only after the basic workflow works.
Provider moderation, model availability and price are controlled by the external services. The plugin does not bypass provider policies or guarantee uncensored output.