Skip to content
Hot eventWatching

PrunaAI releases P-H3-Image image model, supported by ComfyUI and Forge Neo

8 reports6 sourcesUpdated 2 days ago
Subscribe to this event via RSS

Start here

AI overview

PrunaAI has released the image model P-H3-Image based on MiniMax-H3, with open-source weights and native ComfyUI support; PrunaAI claims its generation speed is 5x faster and size is 7x smaller, completing generation in 8 steps, with 4 MP image generation time dropping from 12.4 seconds to 2.6 seconds. The model is a 33.1B-parameter transformer, no CFG required, supports 2-16 MP rendering and editing with up to 8 input images; PrunaAI says it has a built-in Qwen3.5-9B prompt upsampler, while comfyorg says it is equipped with a 4B text encoder. Earlier reports said it was fine-tuned from MiniMax-H3, and later comfyorg said it was distilled from Minimax H3. ComfyUI v0.44.0 has native support, with code merged on 2026-11-03; Aikimi Forge Neo v6.0.0 also adds support, standard model is about 29.8 GB, requires Windows, NVIDIA GPU, Git and internet connection; P-H3 Image and its outputs are subject to MiniMax H3 usage conditions, excluding EU, UK, South Korea, and US. あいきみ tested the model's image editing capability on Forge Neo, on RTX 3090, INT8 body, NVFP4 text encoder, dedicated FP16 VAE, Euler/simple, 6 steps, CFG 1, with editing times of approximately 15 minutes, 4 minutes 35 seconds, and 49 seconds respectively.

Written by AI from the reports · updated 7 days ago

Developments

5 developments

  1. Nov 17 23:13 · 1 report
    Trying the 2step LoRA of MiniMax-H3-Image/P-H3-Image
    note · kongo_jun · 图像视频创作分享:Trying the 2step LoRA of MiniMax-H3-Image and P-H3-Image in ComfyUI
  2. Nov 14 00:22 · 1 report
    Sharing experience running Minimax-H3-Image on a 12GB GPU
    Reddit StableDiffusion · 创作者与工作流分享:Minimax-H3-Image can run on an RTX 3060 12GB + 32GB RAM
  3. Nov 12 17:52 · 1 report
    MiniMax-H3-Image Image Editing and VAE Comparison Evaluation
    あいきみ · 图像编辑与LoRA实测:Testing MiniMax-H3-Image on Aikimi Forge Neo: VAE Comparison
  4. Nov 7 15:46 · 1 report
    Forge Neo v6.0.0 Supports P-H3 Image
    あいきみ · 图像编辑与LoRA实测:Aikimi Forge Neo v6.0.0 Supports P-H3 Image for Image Generation and Editing
  5. Nov 3 08:00 · 4 reports
    PrunaAI releases an image model based on MiniMax-H3
    ComfyUI Wiki · 工作流与开源创作分享:PrunaAI releases the P-H3-Image image model, fine-tuned from MiniMax H3

Report timeline

Follow the reports to see the event from other sides.

Nov 17
  1. note · kongo_jun · 图像视频创作分享
    Trying the 2step LoRA of MiniMax-H3-Image and P-H3-Image in ComfyUI

    The author configured and tested the 2step LoRA of MiniMax-H3-Image and P-H3-Image in ComfyUI; generation times were 16 seconds and 19 seconds respectively, with VRAM at 21GB for both. The author feels that for T2I, P-H3-Image is more to their taste, while for image editing, MiniMax-H3-Image is more to their taste, and notes that the speed-up LoRA seems useful in high-resolution generation.

Nov 14
  1. Reddit StableDiffusion · 创作者与工作流分享
    Minimax-H3-Image can run on an RTX 3060 12GB + 32GB RAM

    A user shared their ComfyUI experience running Minimax-H3-Image (int8 convrot version) on an RTX 3060 12GB + 32GB RAM; it works normally after enabling --disable-dynamic-vram.

Nov 12
  1. あいきみ · 图像编辑与LoRA实测
    Testing MiniMax-H3-Image on Aikimi Forge Neo: VAE Comparison

    Testing MiniMax-H3-Image on Aikimi Forge Neo, verifying image editing capabilities by varying lighting, composition, and art materials. The generation environment was an RTX 3090, INT8 base model, NVFP4 text encoder, dedicated FP16 VAE, Euler/simple, 6 steps, CFG 1, with editing times of approximately 15 minutes, 4 minutes 35 seconds, and 49 seconds, respectively.

Nov 7
  1. あいきみ · 图像编辑与LoRA实测
    Aikimi Forge Neo v6.0.0 Supports P-H3 Image for Image Generation and Editing

    Aikimi Forge Neo update adds support for the P-H3 Image image generation and editing model based on MiniMax H3, available from v6.0.0. It supports text-to-image, image editing, and multi-image reference generation. The standard model is approximately 29.8 GB and requires Windows, an NVIDIA GPU, Git, and an internet connection. P-H3 Image and its outputs are subject to MiniMax H3 usage terms, excluding the EU, UK, South Korea, and the USA.

Nov 4
  1. comfyorg · YouTube 创作视频
    Pruna P-H3-Image is now available in ComfyUI.

    Pruna P-H3-Image is now available in ComfyUI. The model is based on Minimax H3 distillation, equipped with a 4B text encoder, and supports running locally or in the cloud.

  2. Reddit StableDiffusion · 创作者与工作流分享
    Image model P-H3-Image based on MiniMax-H3 released: 7 times smaller, 5 times faster, 8-step generation

    PrunaAI released the image model P-H3-Image based on MiniMax-H3, with open weights and native ComfyUI support, generation speed increased 5 times and volume reduced 7 times. The model only needs 8 steps to generate an image (previously 50 steps), 4 MP image generation time dropped from 12.4 seconds to 2.6 seconds, and it can run on smaller GPUs.

  3. Reddit comfyui · 创作与工作流分享
    PrunaAI releases P-H3-Image, an image model based on MiniMax-H3, 7x smaller and 5x faster

    PrunaAI releases P-H3-Image, an image model based on MiniMax-H3, with open weights and native ComfyUI support. The model is 7x smaller than H3 and 5x faster at generation, completing generation in 8 steps, with a 4 MP image taking only 2.6 seconds. It excels at styles such as realistic photography, documentary and editorial photography, and film stills, and supports 2-16 MP rendering and editing with up to 8 input images.

Nov 3
  1. ComfyUI Wiki · 工作流与开源创作分享
    PrunaAI releases the P-H3-Image image model, fine-tuned from MiniMax H3

    PrunaAI releases the P-H3-Image image model, fine-tuned from MiniMax H3, supporting 8 MP text-to-image generation and editing with up to 8 input images. The model is a 33.1B-parameter transformer, uses 8-step sampling without CFG, includes a built-in Qwen3.5-9B prompt upsampler, and is natively supported in ComfyUI v0.44.0.

Heat trend for this event

Current
3
Peak
14Nov 14 01:00
24h change
-49%

The trend compares only sources observed the whole time, so the range may be smaller than the current heat figure. Hover or tap the chart for each hour. Left and right arrow keys move through it.