v0.1.813-beta v0.1.814-beta
unslothai/unslothSep 22, 2026 → Sep 22, 20260 days between releases
What's New
Adds image editing to Qwen-Image-2.1 and changes default GGUF format to Q4_K_M
Breaking Changes
- Changed default GGUF format from F16/BF16 to Q4_K_M
New Features
- Added image editing support to Qwen-Image-2.1
- Vision models can now see images returned by MCP tools
Improvements
- More reliable training resumes and dataset preparation
- Fixed diffusion black artifacts for A100 and consumer GPUs
Bug Fixes
- Fixed GGUFs not loading for Qwen-Image models
- Fixed diffusers updating issues
Upgrade Notes
The default GGUF format has changed from F16/BF16 to Q4_K_M, which may impact memory usage. Qwen-Image-2.1 now supports both image editing and generation.
The Two Releases
Qwen-Image-2.1 + Skills
Sep 22, 2026· 30 days ago
This release introduces local support for Qwen-Image-2.1 for image generation and adds custom Agent Skills to guide models through specific tasks. It features performance improvements with 2x faster reasoning blocks, enhanced project management with draggable chats, and better Linux/ARM64 support including a native .deb installer and in-app updates.
Full release notesQwen-Image-2.1 + Skills
Sep 22, 2026· 29 days ago
This release enables local execution of Qwen-Image-2.1 with support for both image editing and generation. It introduces custom Agent Skills, enhances chat and project management with a reworked UI, and offers a 2x speed boost for reasoning blocks.
Full release notes