Moss is the best base model we've gotten yet
Yes Krea 2 is amazing but even with loras it still struggles at some things and it has sort of an inflexible style to it.
Proof1 community thread
Key points
- What it struggles with Moss can do right out of the box.
- To me after playing around with it Moss kind of feels like SDXL 2, except it's not only uncensored, the training data included uncensored images.
- For a base model this is huge.
- The finetunes are going to be lit.
Sources (1)
- [1]Moss is the best base model we've gotten yetr/StableDiffusion (top, day) · Oct 17, 07:29 PM
Yes Krea 2 is amazing but even with loras it still struggles at some things and it has sort of an inflexible style to it.
What it struggles with Moss can do right out of the box.
Extractive summary: sentences quoted from the sources.
Before this
- Oct 17, 2026MOSS-Image LoRA training is now in AcademiaSD LoRAlab Trainer Studio (Base and SFT, under 6 GB VRAM)
- Oct 17, 2026Fizgig Lora Isolation - A new process for more powerful LoRAs
- Oct 17, 2026Qwen Flash Next quant and temp comparison - performance vs task time vs quality
- Oct 16, 2026FLUX 3 vs KREA 2 vs Z image turbo vs Qwen 2.1
- Oct 15, 2026What a £300, 35W Snapdragon mini PC can actually do 👀: 35B coder at ~30 tok/s, and an 80B model that doesnt even fit in RAM running at 17 tok/s
- Oct 15, 2026Welcome FLUX.3 Image Klein in Diffusers
