Jan is built by Menlo Research. Its model selection includes DeepSeek alongside other open and online model families. It is a useful option to investigate if you want a choice of models in one interface. The homepage marks its memory feature as coming soon; this profile does not treat it as available.
The quickstart provides installation routes for Mac, Windows and Linux. On first launch, the default model downloads; chatting can begin after completion. The Hub shows hardware fit for additional local models. Download one, then select it in a chat. As a small CurioFuse exercise, ask it to explain a technical term you already know and compare the answer with your own definition. Local models run on your machine; cloud models run on a provider’s servers and require an API key.
For a remote route, the OpenRouter guide requires an OpenRouter key with sufficient credits. Add it under Settings → Model Providers → OpenRouter, then choose a model in a chat. DeepSeek R1 appears among the guide’s model examples. Before choosing a particular model, check its current availability and your access; an example identifier is not an availability guarantee. This route uses the provider connection, rather than the local model workflow.