Over the last few weeks, we've been working on improving Jan's stability. Every update helps us make Jan's experience faster and smoother. With version 0.5.4, you’ll notice AI running quicker on CPU - better performance all around.
Model Downloads
You can now download models directly from the model selector in Threads. No more jumping between different tabs – just pick, download, and get started.
Fast CPU Inference
We've addressed the slower inference speeds on CPU, so you'll notice faster processing times, especially when using larger models.
Model Starts
We tackled the notorious "model can't start / The specified module could not be found" error. Plus, Phi-3 models are now working smoothly even if you're using an outdated version.
Consistent Warnings
Performance warnings are now aligned between Model Hub and Threads, giving you more reliable insights no matter where you're working.
Persistent Thread Settings
Switching between threads used to reset your instruction settings. That’s fixed now! Your settings will stay intact as you jump between old and new threads.
Minor UI Tweaks & Bug Fixes
We’ve also resolved issues with the input slider on the right panel and tackled several smaller bugs to keep everything running smoothly.
For more details, see the GitHub release notes (opens in a new tab).