ComfyUI Releases

Fixes stable release workflow issues and introduces support for Gemma 12B quantized weights while improving LTXV text encoder efficiency.

Jan 8, 2026

A major update adding Sage Attention 3 support, LTXV 2 model support, and various improvements to video generation nodes and API functionality.

Jan 7, 2026

Updates the frontend package and introduces new nodes for manual sigma control and motion, alongside improvements for AMD hardware.

Dec 31, 2025

Introduces a unified jobs API, refactors the Trainer implementation, and adds support for Qwen VAEs and Flux2MaxImage generation.

Dec 24, 2025

Bug fix release focusing on correcting regressions in V3 node processing, adding GPT-Image-1.5 API nodes, and improving Torch import handling.

Dec 18, 2025

Major update introducing support for new video and image models like Kling and Wan2.6, alongside memory optimizations and workflow template updates.

Dec 17, 2025

This release introduces support for the Ovis and Kandinsky5 image models, alongside major VRAM optimizations including temporal rolling VAE and fp16 LoRA processing. It also marks a significant architectural shift by migrating several node files to the V3 schema and updates the API with new features and deprecations.

Dec 10, 2025

A significant update introducing Flux 2 latent preview, Veo3 First-Last-Frame nodes, and improved batch upload UX. Includes VRAM optimizations for Flux 2 and support for the Kling O1 model.

Dec 2, 2025

Introduces the new Z Image model and fixes critical issues with LoRA compatibility on mixed FP8 and Flux 2 memory estimation.

Nov 26, 2025

Focuses on stability and performance, lowering VRAM usage for the Flux 2 text encoder and updating workflow templates.

Nov 26, 2025

Major Flux 2 update introducing the Flux.2 Pro node, Hunyuan latent video nodes, and fixes for API compatibility and text encoder quantization.

Nov 25, 2025

Adds support for the Nano Banana Pro model and Topaz API nodes, alongside improvements to image batching and workflow management.

Nov 21, 2025

Focuses on hardware support for CUDA 12.6 and bug fixes for specific AI models like Hunyuan 3D and EasyCache channel handling.

Nov 19, 2025

Significant performance improvements through default pinned memory usage and VRAM optimization for various models, alongside API client updates.

Nov 18, 2025

Major update introducing a Mixed Precision Quantization System and a new ScaleROPE node, alongside significant speed improvements for offloading.

Nov 5, 2025

Introduces a new endpoint for custom node subgraphs and a major overhaul of the API nodes with async capabilities and V3 schema support.

Oct 28, 2025

This release focuses on performance optimizations, including faster workflow cancellation and memory management improvements for AMD and NVIDIA hardware. It includes updates for Python 3.14 and the frontend, along with continued V3 schema conversions and fixes for API nodes.

Oct 21, 2025

This release heavily prioritizes the migration to the V3 schema, converting core nodes like Flux, SD3, and LoRA extractors. It also includes updates to API nodes (Gemini, Pika) and performance enhancements for AMD hardware.

Oct 14, 2025

This release adds the Sora2 API node and continues the conversion of API nodes and nodes to the V3 schema. It also introduces Gemma 3 as a text encoder.

Oct 8, 2025

This release is characterized by extensive V3 schema conversions for various node types and a shift from soundfile to torchaudio for audio handling. It also introduces linter support for API nodes.

Oct 6, 2025