LLM Providers
LLM Providers in MLJAR Studio
MLJAR Studio supports hosted AI, local desktop model servers, and servers on your own infrastructure. Choose MLJAR AI, OpenAI, OpenRouter, Ollama, LM Studio, Jan, vLLM, llama.cpp, or a custom OpenAI-compatible endpoint for AI-assisted notebooks and data analysis.
Supported LLM providers
MLJAR AI
Built-in hosted AI with account access.
OpenAI
Use your own OpenAI API key and cloud models.
OpenRouter
Choose cloud models from the OpenRouter catalog.
Ollama Local
Run downloaded models with Ollama on your computer.
LM Studio
Serve local models from a desktop interface.
Jan
Connect to the Jan Local API Server.
vLLM
Connect to a model server on your workstation or infrastructure.
llama.cpp
Serve a GGUF chat model with llama-server.
Custom OpenAI-compatible
Connect another compatible server using its API base URL.
Ollama Cloud
Use remote Ollama models without running them locally.
Quick decision table
| Provider | Where the model runs | Setup |
|---|---|---|
| MLJAR AI | Managed cloud | Sign in; availability depends on account access |
| OpenAI | Cloud | API key and model |
| OpenRouter | Cloud | OpenRouter key and model |
| Ollama Local | Local | Ollama and a downloaded model |
| LM Studio | Local or remote server | Start server; optional token |
| Jan | Local or remote server | Start server; Jan server key |
| vLLM | Local or remote server | Serve model; key if required |
| llama.cpp | Local or remote server | Start llama-server; key if required |
| Custom OpenAI-compatible | Depends on endpoint | API URL, format, model, and key if required |
| Ollama Cloud | Cloud | Ollama key and model |
Before configuring a provider
Activate a lifetime MLJAR Studio license to unlock AI Provider Settings. MLJAR AI uses separate account access. For external cloud services, bring the provider API key and model access. For local inference, install the server and download a suitable model first.
How provider setup works
- Open AI Provider Settings and choose the provider by name.
- For local and custom servers, check Base URL and add the server key if required.
- Wait for the available models to load, or use Refresh models, then select a chat-capable model.
- Click Test connection. After it succeeds, click Save provider and confirm the success message.
- Check the active provider in the sidebar and send a short notebook request. Retest after changing provider settings.
For OpenAI-compatible servers, connection testing checks the model list and selected model. A notebook request also verifies generation; agent workflows can require additional model and server support for tool calling.
How to choose
- For a desktop model server, use Ollama, LM Studio, or Jan.
- For a server you manage, use vLLM or llama.cpp.
- For another compatible endpoint, use Custom OpenAI-compatible.
- For hosted models, use MLJAR AI or connect OpenAI, OpenRouter, or Ollama Cloud.
Local inference depends on the actual endpoint and model execution. A remote base URL sends prompts and notebook context to that server, even when you select a local-server preset. Read Local vs Cloud LLMs for the comparison and Troubleshooting for setup errors.
Frequently asked questions
Which local model providers does MLJAR Studio support?
Ollama, LM Studio, Jan, vLLM, llama.cpp, and custom OpenAI-compatible servers. The endpoint and model configuration determine whether inference is local or remote.
Do local providers need an OpenAI API key?
No. Local servers use their own authentication. Most presets allow an empty key when server authentication is disabled. The Jan preset requires the Jan server API key.
Do I need a license to configure providers?
AI Provider Settings require a lifetime MLJAR Studio license. MLJAR hosted AI has separate account access requirements; external providers may have their own usage charges.
Can I switch providers later?
Yes. Select a provider, configure its connection and model, test the connection, and save. Retest whenever you change the settings.