If you're hearing about Jan for the first time: Jan is a desktop app that runs models locally. It's fully free, open-source, and as simple as ChatGPT in UI.
You can now tweak llama.cpp settings, control hardware usage and add any cloud model in Jan. We just released a major update, adding some of the most requested features from local AI communities. Thanks for all the feedback!
New llama.ccp settings
You can now tweak llama.cpp settings directly in Jan's UI.
Also, no more waiting for us to update the Jan to bump the engine - you can now update the engine version yourself.
Settings you can control over:
- llama.cpp backends
- Continuous Batching
- Parallel Operations
- CPU threads
- Flash Attention
- Caching
- KV Cache Type
- mmap
Advanced hardware controls
Hardware control got a big upgrade. You can now activate/deactivate GPUs and see all hardware details in Settings → Hardware.
Remote models update
Managing cloud models is now easier. Instead of manually adding them, you can install custom remote engines via Settings → Engines.
API support for Gemini and DeepSeek is also available.
These updates (and more) are now live in v0.5.15. Update your Jan or grab the latest version here:
- Web: https://jan.ai/
- GitHub: https://github.com/janhq/jan
We'd appreciate all feedback and are happy to hear what you'd like to see next!
Top comments (0)