Changelog 24.12:

Features:

    thp-shrinker: Put max_ptes_none value to 80% for zero filled pages. This will reduce the memory usage for when THP always is used, while maintaining the same performance
    NVIDIA: GSP Firmware gets now automatically disabled, if the users switches on their own to the closed driver
    chwd: NVIDIA: nvidia-powerd services gets enabled for laptops, to reach the most available tdp
    proton: DLSS Frame Generation is now working. This is also expected to work in the future in the upstream proton
    kernel: AMD Cache Optimizer is now applied. Users with dual x3d CCD’s cpus can now switch between having frequency or cache cores preferred
    kernel: amd-pstate: Backported amd-pstate performance fixes for Strix Point
    kernel: Added upstream fixes for the tdp issues on amd rdna2 and rdna3 gpus
    kernel: Added timing fixes for displays with 5120x1440x240 configuration
    kernel: Experimental AutoFDO optimized kernel from Cachy in the repository under “linux-cachyos-autofdo”
    ISO: Added check, if user running handheld edition and warn then, if they are starting the installation on an unsupported device
    ISO: Added check, if the user is using the latest ISO, if not warn them
    Package Updates: linux-cachyos 6.11.7, nvidia 565.57.01, mesa 24.2.6 and more
    KDE: Upgrade KDE 6.2.3-> 6.2.4
    Jan: 0.5.10
    Update Jan’s client request to adapt API changes from Cortex (#4130) 
    Updated Privacy docs (#4080) 
    Increased max-token limit claude-3.5 (#4112)
    Stop word model setting (#4113) 
    Add more Groq models (#3843) 
    Anthropic Model Updates (#3888) 


Bug Fixes:

    refind: partitioning: changed from 3 way partiton layout to 2 way
    netinstall: added kdeplasma-addons to the Plasma installation
    calamares: Fixed a issue, while partitioning with a swap partition
    timesync: Switched from time.google.com to time.cloudflare.com
    Fixed scrollbar issue css (#4191) 
    Model loading gets stuck on stop (#4177) 
    High performance large convo rendering (#4166) 
    Smooth the loader animation (#4167) 
    Glitch theme applies on launch (#4168) 
    Token speed should not be calculated based on state updates (#4159) 
    Remove filter on recommend model (#4156) 
    scroll bottom when switch thread (#4152) 
    convo component re-render every single token yield (#4150) 
    Default local model from starter screen (#4151) 
    App re-render issues caused by bad state handling (#4146) 
    Handle case stop word too long (#4145) 
    Wrong component wrapping and state handling cause excessive app re-render (#4141) 
    Factory reset closes the app & never opens again (#4139) 
    Memory leak lowlight dependencies (#4134) 
    Default local model from starter screen (#4135) 
    Failed to get huggingface models (#4133) 
    gpu mode toggle on by default but not affect for the first time launch (#4132) 
    Active model should be gone as soon as cortex.cpp server is killed (#4131) 
    Invalid request should show clear error message (#4119) 
    Memory leak when the code block response is too long (#4129) 
    Line-number and border color code block UI (#4105) 
    Editing message should not send follow-up messages (#4116) 
    Handle case markdown value no space (#4109) 
    Vulkan responsive (#4099) 
    Reorder alphabetical cloud model on starter screen (#4072) 
    Disabled shortcut create new thread on starter screen (#4074) 
    Responsive vulkan support (#4076) 
    Auto select model when finished download from starter screen (#4083) 

Security updates 

    DSA-5823-1 webkit2gtk security update        
    DSA-5820-1 firefox-esr security update      
    DSA-5818-1 linux security update    
    DSA-5811-1 mpg123 security update   
    DSA-5806-1 libarchive security update       
    DSA-5807-1 nss security update      
    DSA-5800-1 xorg-server security update