models/Precise Flux dev NF4 FP8/FP16 | CLIP and VAE included - v1.0 fp8

Precise Flux dev NF4 FP8/FP16 | CLIP and VAE included - v1.0 fp8

|
8/31/2025
|
11:25:22 PM
| Discussion|
0
Side profile of a human-like android with peeling skin revealing intricate mechanical circuits, gears, and metal plates. Its head is surrounded by abstract floating structures resembling a futuristic city with glowing data streams. The background features a soft beige sky and distant tiny human silhouettes.
Surreal digital illustration showing a beautiful barefoot girl in a gauze dress standing on water with a huge glowing supermoon and dramatic clouds in the background.

Recommended Negative Prompts

ugly, deformed, noisy, blurry, distorted, out of focus

Recommended Parameters

samplers

Euler, Flux Realistic, [Forge] Flux Realistic

steps

20 - 30

cfg

1

resolution

768x1280

vae

VAE - fp8/fp16 mixed

other models

Flux_dev_fp8_nf4 (baafb3faca), flux1-dev-bnb-nf4-v2 (bea01d51bd), flux-dystopian-anime (dedb10e23f13)

Tips

For best results with lors in Forge, select Diffusion in Low Bits as Automatic (fp16 LoRA).

In Forge, use Euler or Flux Realistic samplers with Schedule type: Simple.

Version Highlights

This model is converted to bnb nf4 by mixing Flux dev fp8 with VAE, text encoder T5xxl-fp8 and Clip-l. All included!

Creator Sponsors

Generated outputs can be used for personal, scientific, and commercial purposes as described in the flux-1-dev-non-commercial-license

This model is converted to bnb nf4 by mixing Flux dev fp8 with VAE, text encoder T5xxl-fp8 and Clip-l. All included!

For those who don't care about size, also blended full dev fp16 with VAE, text encoder T5xxl-fp16 and Clip-l in NF4. All included!

From many observations, the FP8 is a bit more accurate in understanding the prompt. And having assembled this build and tested it, I came to the same conclusion.

This model is a bit more accurate than the familiar NF4 v2. I've done a lot of generation with complex prompts to assert this. And I am completely satisfied with the way this model understands me. With simple prompts I didn't notice any difference.

The model works well with lors in Forge, in which I create in. Don't forget to select in Diffusion in Low Bits: Automatic (fp16 LoRA).

In Forge I choose Euler or Flux Realistic samplers, Schedule type: Simple. CFG: 1, number of steps 20-30.

Generated outputs can be used for personal, scientific, and commercial purposes as described in the flux-1-dev-non-commercial-license

Previous
Stabilizer IL/NAI - illus01 v1.165c
Next
Flux chaotic expressive lineart - v1.0

Model Details

Model type

Checkpoint

Base model

Flux.1 D

Model version

v1.0 fp8

Model hash

46dacf0e3b

Creator

Discussion

Please log in to leave a comment.

Model Collection - Precise Flux dev NF4 FP8/FP16 | CLIP and VAE included

Images by Precise Flux dev NF4 FP8/FP16 | CLIP and VAE included - v1.0 fp8

base model Images

fantasy Images

photorealistic Images

realistic Images