Jan Version 0.5.8

Major Changes

Core Engine Update

• Migrated from llama.cpp (Nitro) to llama.cpp (Cortex) as Jan's core engine

Note about model compatibility from v0.5.8:

When you update to v0.5.8 > Download models in this version > Then downgrade to versions before v0.5.8:
- Some models may temporarily not appear in the app due to configuration format changes (model.json → model.yaml), while the model files remain safe and are not deleted
- You can re-import your existing models after downgrading
When upgrading back to 0.5.8+, all your models will be compatible again

Version Management Enhancement

• Separated installation channels: Stable, Beta, and Nightly versions can now be installed independently
• Important: Existing Nightly users will be automatically migrated to Stable (from v0.5.8)
• New Nightly builds will be purely for testing purposes, primarily for internal team and contributors

UI/UX Improvements

• Added model ID display in Local API Server
• Reduced animation effects for better performance
• Auto-focus on chat input box when opening threads
• Reduced aggressive blur effect on modals for better visibility
• Enhanced system monitor interface
• Expanded clickable areas of icon button for better interaction
• Improved text alignment for model lists
• Fixed inconsistent left panel background in responsive mode
• Fixed various edge cases with chat input behavior & syntax highlighting
• Fixed slate React version compatibility

Other Improvements

• Removed deprecated model GPT-4 with Vision (Preview)
• Added model Qwen2.5 Coder (14b/32b)
• CUDA dependencies are now included by default for improved onboarding experience