v0.42.1 v0.43.0
Comfy-Org/ComfyUIOct 19, 2026 → Oct 27, 20268 days between releases
What's New
v0.43.0 is a broad feature release adding MOSS-image and FIBO 2.0 model support, many Partner Nodes, and core performance work on top of the v0.42.1 crash fix.
New Features
- MOSS-image gains RoPE scaling, color-depth control, edit LoRA support, attention-bias conditioning, image-strength controls, and a model-specific merge node.
- FIBO 2.0 support added with NTK-aware RoPE scaling and corrected negative-prompt handling.
- New core capabilities include a Branch node, Pad Image node, ACEScg color-space support, and a cloud endpoint deployment feature flag.
- Partner Nodes expanded across HeyGen, ClarityAI, BFL FLUX 3, Bria, Krea, VEED, Mirelo, Meshy, PixVerse, Ideogram, and Kling.
- API Partner Nodes now renew sign-in tokens automatically.
Improvements
- Performance and compatibility work landed for Qwen 3.x models.
- Dynamic VRAM KV caches introduced to improve memory utilization.
- AMD VAE attention path optimized.
- Flux2 VAE fusion handling improved.
- Model prefetch now synchronizes all graph cleanups for more predictable memory behavior.
Bug Fixes
- Corrected the top-p sampling threshold.
- Fixed negative-prompt handling for FIBO 2.0 workflows.
Upgrade Notes
This is a minor version bump with no documented breaking changes, so a standard upgrade should suffice. Bundled workflow templates moved to v0.11.92, so re-check any customized templates after upgrading. The top-p sampling correction may subtly change generation output, so re-validate tuned sampling settings.
The Two Releases
v0.42.1
Oct 19, 2026· 9 days ago
This release addresses stability issues where specific users encountered crashes while utilizing the MOSS image template workflow. The update aims to improve the reliability of the software for those working with complex image generation setups.
Full release notesv0.43.0
Oct 27, 2026· 1 day ago
ComfyUI v0.43.0 is a broad feature release adding expanded MOSS-image and FIBO 2.0 model support, many new Partner Nodes, and core workflow improvements. It also includes performance work for Qwen 3.x, dynamic VRAM KV caches, AMD VAE attention, and Flux2 VAE handling. New utility nodes and cloud/API-related additions round out the release.
Full release notes