v0.1.906-beta v0.1.907-beta
unslothai/unslothOct 12, 2026 → Oct 13, 20261 day between releases
What's New
v0.1.907-beta adds major UI performance gains, new video models, and expanded browser automation.
New Features
- Mouse movement and drag capabilities in browser automation
- Support for LTX-2.5 video generation and Wan2.2 image-to-video
- Fused AdamW for image LoRA training on NVIDIA GPUs
- FP8/MXFP8 support for LoRA training on Ada, Hopper, and Blackwell GPUs
- NVIDIA RTX Spark hardware support
Improvements
- Up to 10x faster UI/UX and 5x faster inference with improved cache hit rates
- Better compaction for all APIs and improved UI handling
- Faster startup data usage and memory for long training histories
- Improved compatibility for Windows AMD, Intel XPU, and CUDA/PyTorch
- Faster prefill and speculative decoding for Qwen3.5, Qwen3.6, and Qwen3.8
Bug Fixes
- Fixed Windows AMD, Intel XPU, and CUDA/PyTorch compatibility issues
- Resolved slowdown issues and MoE Strata style problems
- Restored desktop drag-and-drop file support
- Improved file handling for desktop environments
- Fixed issues with long GGUF conversations and tool calls
Upgrade Notes
Expect significant speed improvements in the UI and inference. New video generation models (LTX-2.5, Wan2.2) are available. Browser automation now supports mouse movement, dragging, and scripts.
The Two Releases
Browser Use + Qwen-Image-2.1-Turbo
Oct 12, 2026· 10 days ago
This release introduces browser automation capabilities for models, allowing them to interact with web pages directly. It also brings significant performance improvements for image generation and training, specifically with the Qwen-Image-2.1-Turbo model and speculative decoding on Mac. Additionally, it adds support for text-diffusion LLM training and faster MoE GGUFs.
Full release notesLarge Performance Gains + More Browser Use
Oct 13, 2026· 9 days ago
This release focuses on significant performance improvements across the UI, inference, and training, alongside expanded browser automation capabilities and support for new video/image generation models like LTX-2.5 and Wan2.2.
Full release notes