Installer Update: Faster Dependencies, Qwen Support & What’s Next ⚡
Added 2026-01-20 19:18:07 +0000 UTCHey Friends,
The custom installer just got a major quality-of-life upgrade! 🚀
I’ve implemented a new Python dependency resolver (uv), which means your installations and updates are going to be significantly faster and smoother. You have to activate it under Experimental settings.
New Support Added:
AI-Toolkit
Forge
Qwen Trio: Full support for Qwen 2511, 2512, and Qwen Layered.
If you aren't sure which Qwen model does what (or why the "Layered" model is a game-changer), check out my latest AI Gazette video where I break it all down!
Coming Soon:
I’m currently working on adding support for Flux Klein and LTX 2. Give me a couple of days to get those dialled in for you. The auto-updater is also still on the roadmap and will be part of a future release.
Thanks for the support!
Best,
Chris
Comments
***Edited Response*** I had to install UV and once installed, the installer was able to install everything without issue. Thank you. *** did start getting this when running: Device: cuda:0 NVIDIA GeForce RTX 5060 Ti : cudaMallocAsync Using async weight offloading with 2 streams Enabled pinned memory 36583.0 WARNING: You need pytorch with cu130 or higher to use optimized CUDA operations. Found comfy_kitchen backend eager: {'available': True, 'disabled': False, 'unavailable_reason': None, 'capabilities': ['apply_rope', 'apply_rope1', 'dequantize_nvfp4', 'dequantize_per_tensor_fp8', 'quantize_nvfp4', 'quantize_per_tensor_fp8', 'scaled_mm_nvfp4']} Found comfy_kitchen backend triton: {'available': True, 'disabled': True, 'unavailable_reason': None, 'capabilities': ['apply_rope', 'apply_rope1', 'dequantize_nvfp4', 'dequantize_per_tensor_fp8', 'quantize_nvfp4', 'quantize_per_tensor_fp8']} Found comfy_kitchen backend cuda: {'available': True, 'disabled': True, 'unavailable_reason': None, 'capabilities': ['apply_rope', 'apply_rope1', 'dequantize_nvfp4', 'dequantize_per_tensor_fp8', 'quantize_nvfp4', 'quantize_per_tensor_fp8']} Using pytorch attention Python version: 3.12.10 (tags/v3.12.10:0cc8128, Apr 8 2025, 12:21:36) [MSC v.1943 64 bit (AMD64)] ComfyUI version: 0.10.0 ComfyUI frontend version: 1.37.11
James Charleston II
2026-01-20 23:05:24 +0000 UTCThank you!
David Thielen
2026-01-20 19:31:13 +0000 UTC
