BibGenie DocsModels
Understand official models, custom models, the AUTO model, credit multipliers, and model refresh behavior.
BibGenie supports two kinds of models: official models and custom models.
- Official models are maintained by BibGenie and work out of the box. You do not need to configure your own API key.
- Custom models are configured by you, including the provider, Base URL, API key, and model capabilities.
You can manage both from Model Settings.
Official Models
There are many AI models on the market, but only a small subset works well as a Zotero agent. BibGenie does not choose models based only on public benchmarks. We care about whether they can reliably handle real research workflows:
- Long-context paper reading
- Multi-paper comparison
- Complex research reasoning
- Tool calling and structured output
- Multimodal inputs such as figures, formulas, and screenshots
- Accurate answers grounded in Zotero context
Provider configuration also matters. The same model can behave very differently across providers because of context limits, tool-calling support, rate limits, parameter defaults, and network reliability.
To reduce configuration work for users, BibGenie does several things:
We test a selected set of models on paper reading, literature search, library management, and note-taking workflows.
We discuss recommended deployment and usage patterns with model or provider teams when possible.
We work with selected providers to make sure models are served in a way that works well for BibGenie.
We benchmark model-provider combinations and maintain a recommendation list.
Custom Models
If you have your own API key or want to use a local model, you can add a custom model.
Custom models are useful when:
- You already pay for a provider such as OpenAI, Anthropic, DeepSeek, or OpenRouter.
- Your institution requires requests to go directly to a specific provider.
- You want to use Ollama or LM Studio locally.
- You need a model that is not currently included in the official list.
See Custom Models for setup instructions.
The AUTO Model
BibGenie includes a permanently free AUTO model. It requires no setup and is suitable for lightweight tasks:
- Simple translation
- Short text rewriting
- Basic Q&A
- Trying out the BibGenie experience
AUTO is not designed for complex work
The AUTO model is less capable than recommended official models. We do not recommend it for long PDF deep reading, complex reasoning, multi-paper literature reviews, library organization, or tasks that require high accuracy.
What Do the Multipliers Mean?
You may see labels such as 0x, 1.5x, or 2x after a model name.
These labels indicate the approximate speed at which the model may consume credits:
| Label | Meaning |
|---|---|
0x | Usually means the model consumes almost no credits or the cost is covered by the system, depending on the current product policy. |
1x | Standard consumption speed. |
1.5x | May consume credits faster than a standard model for a similar task. |
2x or higher | Usually indicates a higher-cost model, best reserved for more demanding tasks. |
Actual usage depends on tokens and model pricing
Multipliers are shown as a user-facing hint. Actual credit usage is calculated from input tokens, output tokens, model pricing, and tool usage.
Refreshing the Model List
The official model list refreshes automatically. You can also refresh it manually:
Open BibGenie.
Go to Model Settings.
Click Refresh in the official models section.
If a model is missing, possible reasons include:
- Your plugin version is out of date.
- The model list has not refreshed yet.
- The model is temporarily unavailable or being adjusted.
- Your account status has not synced.
Choosing a Model
Translation and short Q&A
Start with AUTO or a low-multiplier model.
Single-paper summaries
Use an official model with reliable context handling and strong output quality.
Complex reasoning and reviews
Use a model with stronger reasoning ability and a longer context window.
Privacy or local execution
Use BYOK, a custom provider, Ollama, or LM Studio.