🚀 Highlights

1. Engine Improvements

• Revamped support for remote engines (OpenAI, Anthropic, etc.)
• Allow users to install custom remote engines and add any models, which means they can connect & use any cloud models they prefer
  See docs: https://jan.ai/docs/install-engines
• Better local engine settings & management with version and variant selection

2. Improved Hardware Controls

• We migrated Jan's monitoring and hardware compatibility APIs to Cortex
• Enhanced UI of hardware settings

🚀 New models

• Added Gemini API
• Added DeepSeek API
• Added DeepSeek R1 Distill models to Hub

🐛 Minor Bug Fixes & Enhancement

• Fix OpenAI's o1 & o3 models support
• Resolved "missing model" error in Threads screen with cloud models, which causes missing starter screen error
• Resolved some edge cases around model loading error
• Fixed incorrect default max_tokens setting
• Fixed empty response issues with Claude models on regeneration
• Fixed ngl mismatch between UI and model configuration options
• Fixed issue where the app fails to load models on Windows due to the absence of CUDA
• Fixed broken UI of file & image upload
  And other minor bug fixes...

📚 Enhancements & Documentation

• Improved the whole Jan documentation
• Improved in-app copy and descriptions
• Standardized input field behavior
• Improved UI of error notification
• Better UI & animation for reasoning models while thinking
• Better vertical responsiveness when open System Monitor across app
• Threads now default to user's last chosen model when they create a new thread
  And many minor improvements...