## Comfy-Org/ComfyUI — v0.37.4…v0.38.0

_59 commits._

### Features
- **Support w6a8 quantization format (#16483)** (491bf3b)
- **Support Qwen-Image 2.1 union fun controlnet (#16519)** (3b3939a)
- **Add comfy_attention and AttentionTensorContainer to a few models. (#16595)** (8d53494)
- **Support ID-V2V (#15139)** (5c4d256)
- **Support tiny VAE for Qwen-Image 2.1 (#16552)** (a73d24b)
- **Add HDR LogC3 and ACEScct to Convert Color Space node. (#16541)** (8308c3b)
- **[Partner Nodes] feat(ByteDance): add Seedance 2.5 Draft mode (#16529)** (9f93254)
- **feat(npu): support async weight offload streams (#16057)** (78368ea)
- **Add missing RDNA2 arches to list. (#16537)** (ca3c529)
- **[Partner Nodes] feat(ByteDance): add Seedream 5.0 Flash (#16522)** (9381048)
- **[Partner Nodes] feat(OpenAI): add GPT-6 Sol and Luna models (#16520)** (6a103b1)
- **Lower memory usage and .comfy_attention support for lumina family models. (#16515)** (1568e6c)
- **feat: ming-image support (#16482)** (3b4c0b0)
- **[Partner Nodes] feat(Recraft): add Recraft V4.1 Flash to the V4 text-to-image node (#16501)** (15ef24d)
- **[Partner Nodes] feat(Anthropic): add Claude Opus 5.5 to the Claude node (#16479)** (d435c29)
- **[Partner Nodes] feat(Quiver): add Arrow 2 models with reasoning effort to the SVG nodes (#16478)** (5123892)
- **support union cn 2.0 (#16471)** (95539f5)
- **TextGenerate: Add system_prompt input and separate thinking output (CORE-460) (#16442)** (9498c42)
- **[Partner Nodes] feat(Tencent): add Hunyuan Image 3.5 text-to-image and edit nodes (#16462)** (b09760d)

### Fixes
- **perf(assets): keep the UI responsive during asset scans and rescan output by listing folders (#16546)** (56c5005)
- **fix(assets): Start without the assets packages; --enable-assets reports what is missing and runs with assets disabled (#16580)** (d669bcf)
- **fix(db): wait briefly for the database lock at startup (#16602)** (a35fcba)
- **Fix potential regression with previous PR. (#16596)** (4ef23c3)
- **perf(assets): replace pathlib prefix checks in the asset scan (#16543)** (7897b4e)
- **Fix HDR output issue with negative values in Convert Image Color Space. (#16565)** (f2a419e)
- **Fix Qwen VL image preprocessing crash on 4-channel (RGBA) input (#16548)** (39cb5a8)
- **Small fix: skip unnecessary work. (#16540)** (88ab4a0)
- **fix(assets): harden the asset catalogue's failure paths (#16393)** (02ba106)
- **Fix saving/loading ming image model breaking detection. (#16514)** (41c83ee)
- **Fix ImageUpscaleWithModel crashing on RGBA images (#16500)** (557597f)
- **Fix MiniMax-H3 VAE rms_rope crash on offloaded qk_norm_scale (#16485)** (912fca4)

### Backend
- **ComfyUI v0.38.0** (6b747c0)
- **Update codeowners. (#16655)** (7a8f3ad)
- **Avoid unnecessary write. (#16639)** (a716932)
- **Speedup generation on qwen3.5/3.8 with long contexts. (#16638)** (67f11cc)
- **Fixes for last PR. (#16637)** (9671d89)
- **Optimize Seedvr2 (#16530)** (84131a6)
- **Update comfy-kitchen package version to 0.2.36 (#16635)** (6f6a274)
- **Searching "kitchen" now returns the attention node. (#16632)** (2139131)
- **Speedup Yue2 AR (#16626)** (7fbcfa8)
- **Clamp Qwen Image 2.1 fp16 activations in place. (#16608)** (47e63a2)
- **[Partner Nodes] chore(Sora): remove deprecated nodes (#16609)** (a9b777b)
- **Mention some ComfyUI optimization details in README. (#16576)** (79be670)
- **Update sheetsage2 abc creation from upstream yue code. (#16569)** (66b68a3)
- **Move file reads out of the remaining asset write transactions (#16486)** (f427c3a)
- **Take the SQLite write lock up front for asset scan and output-registration writes (#16480)** (9d3bfc3)
- **Port some optimizations to flux model family. (#16488)** (b5cc883)
- **Fast disk detection to all model loaders. (#16425)** (e638023)
- **Blend H3 VAE tiles against composited neighbours (#16436)** (fc584aa)
- **Cuda graphs + memory compiler on ace step 1.5 autoregressive model. (#16461)** (d158420)
- **Remove torchaudio dependency. (#16457)** (b16023b)
- **JsonExtractString can now handle JSON with prefix and/or suffix. (#16439)** (b0f4b7b)
- **Allow model files to contain which attention should be used for each block. (#16419)** (c194dd0)
- **Compile Qwen Image 2.1 transformer blocks (#16430)** (1d61dcc)
- **Improve qwen 2.1 KV cache location logic (#16429)** (2f7c6d4)
- **Bump back frontend to 1.53.6 after dowgrading it on the stable commit.** (0f74f7f)

### Chore
- **chore: update workflow templates to v0.11.70 (#16557)** (f02c145)
- **chore: update workflow templates to v0.11.69 (#16503)** (2255709)
- **chore: update workflow templates to v0.11.68 (#16466)** (b33e2b5)

_Recap by [Repo Wrapped](https://repowrapped.com/gh/Comfy-Org/ComfyUI?utm_source=github-action)._