DEV Community

AI-KSK
AI-KSK

Posted on Originally published at huggingface.co

Qwen Image: MiniMax Larry Sageattention — 11 Workflows, Read From the Graphs

Qwen Image: MiniMax Larry Sageattention — 11 Workflows, Read From the Graphs

This release ships 11 workflow files built around Qwen Image. Together they total 419 nodes. Rather than summarise the video, this write-up reads the graphs directly — what each one loads, how big its latents are, and which sampler it runs.

The workflows at a glance

Workflow Nodes Links Sampler Latent
MiniMax H3 Larry + Sageattention加速 91 111 euler —
MiniMax+H3+Larry+attention+cache+满血加速版 71 86 euler —
MiniMax H3 INT8 W4A8 加速 62 78 res_multistep —
MiniMax+H3参考生视频6倍加速80GB满配版 (1) 27 34 euler —
MiniMax H3参考生视频6倍加速24GB通用版 27 34 euler —
MiniMax H3参考生视频6倍加速48GB高配版 (1) 27 34 euler —
+MiniMax+H3图生视频6倍加速24GB通用版 24 29 euler —
MiniMax+H3图生视频6倍加速80GB满配版 24 29 euler —
MiniMax H3图生视频6倍加速48GB高配版 (1) 24 29 euler —
MiniMax H3 Larry Turbo 加速 21 27 euler —
MiniMax H3 LightX2V Turbo 加速 21 27 er_sde —

The node count range is wide: MiniMax H3 Larry Turbo 加速 at 21 nodes against MiniMax H3 Larry + Sageattention加速 at 91. That gap is usually the difference between a single-pass preview graph and one that adds conditioning, upscale stages, or a second sampling pass.

What the graphs load

Model file Loader node
minimax_h3_video_vae_fp16.safetensors VAELoader
minimax_h3_audio_vae_fp32.safetensors VAELoader
qwen3vl_32b_minimax_h3_bf16.safetensors CLIPLoader
minimax_h3_fl2va_bf16.safetensors UNETLoader
qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors CLIPLoader
minimax_h3_video_vae_int8_convrot.safetensors VAELoader
minimax_h3_ref2va_pruned_int8_convrot.safetensors UNETLoader
minimax_h3_fl2va_pruned_int8_convrot.safetensors UNETLoader
minimax_h3_fl2va_pruned_w4a8_mixed.safetensors UNETLoader
minimax_h3_fl2va_int8_convrot.safetensors UNETLoader
qwen3vl_32b_minimax_h3_int8_convrot.safetensors CLIPLoader
minimax_h3_ref2va_bf16.safetensors UNETLoader
minimax_h3_ref2va_int8_convrot.safetensors UNETLoader

LoRAs in play

LoRA Weight
minimax_h3_turbo_v4_step600_ema_pruned_comfyui.safetensors 1
minimax_h3_fl2v_lightx2v_turbo_4step_v0.1_comfy.safetensors 0.7500000000000001

Not every LoRA runs at 1.0 — the non-default weights here are [0.75]. A weight above 1.0 amplifies the LoRA's effect past what its author trained for; below 1.0 backs it off. Both are deliberate tuning choices, not leftovers.

Sampler choice

Samplers used: er_sde, euler, res_multistep.

A plain euler is deterministic — good when the constraint is the point and noise per step would only perturb the trajectory. Ancestral variants inject noise each step and are the right pick when you want the model to explore rather than commit.

Where this set is weak

  1. Node count is not quality. The largest graph is not automatically the best one; a 40-node graph with a good prompt beats a 200-node graph with a bad one.
  2. No measured comparison in the files. The workflows carry settings, not benchmarks. Any quality claim needs your own A/B run.
  3. Model files are not included. You supply the weights; the graphs only reference filenames.
  4. Version drift. A graph pinned to a specific checkpoint revision will run on a newer one but may produce different output.

Practical order

Open the smallest graph first. It has the fewest moving parts and will confirm your models load before you spend time on the elaborate ones.

Workflows

11 graphs published on RunningHub with invite code rh-v1111:

  • +MiniMax+H3图生视频6倍加速24GB通用版
  • MiniMax+H3参考生视频6倍加速80GB满配版 (1)
  • MiniMax+H3图生视频6倍加速80GB满配版
  • MiniMax H3 INT8 W4A8 加速
  • MiniMax H3 Larry + Sageattention加速
  • MiniMax H3 Larry Turbo 加速
  • MiniMax H3 LightX2V Turbo 加速
  • MiniMax H3参考生视频6倍加速24GB通用版
  • MiniMax H3参考生视频6倍加速48GB高配版 (1)
  • MiniMax H3图生视频6倍加速48GB高配版 (1)
  • MiniMax+H3+Larry+attention+cache+满血加速版

Files

https://huggingface.co/AI-KSK/ksk-video-291-comfyui-workflows

Video walkthrough: https://youtu.be/SUJkkT6XLHI
Bilibili: https://www.bilibili.com/video/BV17nuU6xE3g/

Notes

  • Workflow files only. Model weights are not included and each keeps its own licence.
  • Node counts, geometries and loader filenames above are read from the published JSON files.
  • Independent AI-KSK release. Not affiliated with the model authors or RunningHub.

Optional API platform - 88API (affiliate link)

Explore 88API for access to multiple AI model APIs. Available models, routes and prices are listed on the platform; Alipay top-ups are supported.

https://88api.ai/sign-up?aff=usx3

Disclosure: I may earn a commission from eligible purchases after you register through this referral link.

Top comments (0)