PaperLabUse the model you trust, local or cloud
PaperLab doesn’t lock you to one AI provider. Add connections in Settings: a local Ollama model, Anthropic with your own key, or any OpenAI-compatible server, including OpenAI, OpenRouter, Groq, Mistral, DeepSeek, Gemini, LM Studio, vLLM and llama.cpp. Pick the model per question in the chat panel. Cloud models are tagged “Cloud”, and every answer keeps the model and connection that wrote it.
What you can do
- Connect Ollama, Anthropic or any OpenAI-compatible server.
- Test a connection, choose which models chat lists, and set the default.
- Pull and delete Ollama models from Settings, with live progress.
- Pick the model per question from the chat panel.
- Cloud models are tagged “Cloud”; local ones keep your text on your machine.
- Keys stay in your local database and are never sent back to the browser.
Tips
- Try qwen3:8b on Ollama as a first local model on a 16 GB laptop.
- Use a cloud model for hard questions on published papers, and a local one for unpublished work.
Questions people ask
Which local models work?
Any model Ollama, LM Studio, vLLM or llama.cpp serves. Small models around 7–8 billion parameters run on a 16 GB laptop.
Where are API keys stored?
In your local database. They are never sent back to the browser.
Do I need a model at all?
No. Reading, highlighting and notes work without one. Chat and suggestions need one.
Works well with
Guides
- How to chat with a PDF locally, without uploading it — Ask papers questions with a local model, and check every answer against the page.
- Open-source alternative to ChatPDF, Elicit and SciSpace — Online AI paper tools compared with a local, open-source option.
- Running a local LLM for research — Hardware, Ollama, choosing a model, and making it answer well.