GitHub
20 articles in the TruthFoundry index. Each links out to the original.
- Unsloth Releases Update With Decision Model Training And ComfyUI Support · Unsloth launches update enabling Jev-style decision model training for all LLMs.
- Unsloth Desktop Release Brings Faster Laya Decisions and 4-Bit Training · Unsloth Desktop updates navigation, error clarity, and Laya decision speeds while optimizing 4-bit model training.
- Unsloth Adds Decision Models, Faster Image Generation, and Apple Silicon Support · Unsloth releases Decision Models, a unified Library, and significant performance boosts for Apple Silicon and image generation.
- Unsloth releases prebuilt binaries for Flash-Attention2, Causal-Conv1D, Mamba SSM · Unsloth published prebuilt CUDA 13 ML library wheels on 27 September.
- Unsloth Releases Qwen-Image-2.1 with Custom Agent Skills and 2x Faster Reasoning · Unsloth launches Qwen-Image-2.1 locally with new agent skills, faster reasoning, and improved Linux support.
- Unsloth Releases Updated Docker with AMD, Multi-User, and Windows ARM64 Support · Unsloth launches new Docker image featuring AMD ROCm support, multi-user isolation, and Windows ARM64 CUDA capabilities.
- Unsloth releases Windows ARM64 binaries via GitHub · Unsloth published Windows ARM64 build binaries in a GitHub release on 16 September.
- Unsloth Releases Major Update with 1.7x Faster Diffusion and 250+ Bug Fixes · Unsloth releases a large update featuring significant performance gains and over 250 bug fixes for its AI software.
- Unsloth Releases Performance Update with Vulkan Support, PyTorch 2.11, and Over 200 Bug Fixes · Unsloth's new release adds default Vulkan support for AMD GPUs, updates PyTorch to 2.11, and includes over 200 bug fixes.
- Unsloth Releases Qwen3.8-Flash and GLM-5.3-Flash with 2x Speed Boost · Unsloth launches Qwen3.8-Flash and GLM-5.3-Flash models delivering up to 2x faster inference via MTP.
- Unsloth Desktop Update Enables 2x Faster Qwen3.8-Flash and GLM-5.3-Flash via MTP · Unsloth AI's new release enables MTP by default for Qwen3.8-Flash and GLM-5.3-Flash, claiming up to 2x faster generation.
- Unsloth adds local support for Qwen3.8-Flash-Next and GLM-5.3-Flash models · Unsloth's new release enables running 125B Qwen3.8-Flash-Next and 320B GLM-5.3-Flash models locally on 75GB and 102GB memory.
- Unsloth releases bug fixes, auto compaction, LAN remote access · Unsloth release adds experimental auto compaction, LAN remote access, MLX/AMD fixes, and 170+ bug fixes.
- Unsloth Desktop Release Adds Auto Compaction and LAN Remote Access · Unsloth Desktop introduces experimental auto compaction for long chats and preview LAN remote access features.
- Unsloth merges main branch with studio-mmproj-fit updates · Unsloth merges main branch into studio-mmproj-fit, resolving frontend import errors.
- Unsloth Desktop Adds Support for Qwen3.8 and MiniMax-H3 Models · Unsloth Desktop now allows local running and fine-tuning of Qwen3.8 and MiniMax-H3 models.
- Unsloth Releases v0.1.71-Beta with Media Picker and H3 Updates · Unsloth AI releases v0.1.71-beta featuring media picker constraints and H3 naming fixes.
- Unsloth Desktop Launches Open-Source AI Training App for Windows, Mac, Linux · Unsloth releases a free, open-source desktop application for local AI model training and inference across major operating systems.
- Unsloth launches free open-source desktop app for local AI model training and deployment · Unsloth Desktop is a free open-source app for running and training AI models locally on Windows, macOS, and Linux.
- Unsloth v0.1.62-beta released with bug fixes, 488 commits listed · danielhanchen released Unsloth v0.1.62-beta on 11 August, describing it as containing many bug fixes.