Online design platform Freepik has launched a new generative AI image model called F Lite, trained exclusively on commercially licensed and safe-for-work images.
Developed in partnership with AI startup Fal.ai, F Lite was trained over two months using 64 Nvidia H100 GPUs and features approximately 10 billion parameters.
In a market where copyright concerns around AI training data are intensifying, F Lite stands out by being trained on an internal dataset of around 80 million images—setting it apart from models trained on web-scraped content. Freepik says the model aims to offer developers an open, customizable solution without relying on potentially infringing data sources.
We’ve been secretly working on this for months! It feels good to finally share it!
— Javi Lopez ⛩️ (@javilopen) April 29, 2025
LINKS:
• Regular version: more predictable and prompt-faithful, but less artistic: https://t.co/MyWsKer9Ir
• Texture version: is more chaotic and error-prone, but delivers better textures and… pic.twitter.com/GX5mIpYE8O
F Lite comes in two variants:
- Standard – More predictable and faithful to prompts.
- Texture – Offers richer compositions and textures but is slightly more prone to errors.
While Freepik doesn’t claim that F Lite outperforms models like Midjourney V7 or Flux by Black Forest Labs, the focus is on openness and ethical data sourcing. However, running the model requires serious hardware—a GPU with at least 24GB of VRAM.
As legal battles around AI-generated content continue, Freepik joins companies like Adobe, Shutterstock, Getty Images, and Moonvalley in championing licensed-data training, potentially shaping the future of ethical AI image generation.
If this article helped you, please consider supporting our work. Every small contribution keeps Abijita.com independent and running.
