This update brings significant improvements to model handling, user interface enhancements, and various bug fixes to improve your experience with Jan. Here's what's new:

New models & Enhancement

• Added llama3.2 and qwen2.5 models.
• Improved Jan's new starter screen with subsequent refinements to onboard new users.
• Added tabs in model selection to help navigate between on-device & cloud models better.
• Enhanced text area auto-resize functionality.

Bug Fixes:

• Corrected GGUF model metadata handling and downloading, which addressed the bug that prevented users from using models imported directly from Hugging Face and caused the application not to be able to utilize the user's GPU correctly.
• Improved models and thread caching for better performance.
• Resolved XML rendering issues.
• Corrected model dropdown to show recommended models for download and fixed local model display.
• Addressed the "Unable to use TensorRT-LLM" issue.
• Fixed an issue where failed download models were still being used.
• Fixed API request body size limits.
• Resolved problems with instruction handling for different models, including Claude 3.5.
• Fixed UI issues such as input text overlapping model settings and toolbar overlap with chat input.
• Fixed some issues with model running & other small issues...