.jpg)
Bria has launched its third-generation text-to-image AI model, a fully open-source solution that matches industry-leading performance with just 4 billion parameters. Built entirely on licensed data, this model offers enterprises a responsible and efficient AI solution for visual content creation, setting a new standard in the generative AI landscape.
Bria releases first fully open-source text-to-image AI model.
Matches top models with 66% fewer parameters, using 4B parameters.
Trained exclusively on licensed data, ensuring ethical AI development.
Twice as easy to fine-tune, requiring 50% less compute and data.
Available on Hugging Face with tools like ControlNets and IP adaptors.
Supports enterprise needs with content moderation and security features.
Bria’s third-generation text-to-image model, now open-source, delivers performance comparable to leading models while using 66% fewer parameters. With only 4 billion parameters, it achieves efficiency without compromising quality, as validated by a study involving thousands of design professionals. The model excels in aesthetic quality, prompt adherence, text rendering accuracy, and fine-tuning effectiveness, making it a game-changer for enterprise AI adoption.
“While the industry races to build ever-larger models using scraped web data, we’ve proven that smaller, ethically-trained models can deliver equivalent performance,” said Yair Adato, CEO of Bria.
Unlike competitors that rely on web-scraped data, Bria’s model is trained solely on licensed content, mitigating copyright and trademark risks. This approach ensures fair compensation for artists, photographers, and content creators, fostering a sustainable creative ecosystem. By prioritizing ethical data sourcing, Bria enables businesses to adopt AI confidently, free from legal concerns.
“Every image generated by Bria represents a vote for a sustainable creative ecosystem,” added Vered Horesh, CSO at Bria.
Available immediately on Hugging Face, Bria’s model includes a comprehensive development framework with ControlNets, IP adaptors, and auxiliary models. Its platform-as-a-service offering optimizes developer velocity with built-in content moderation, enterprise-grade security, and support for multiple access methods, including MCP servers, Figma plugins, and Adobe Creative Suite integration. The model’s fine-tuning process is twice as efficient, requiring 50% less compute and data, making it ideal for enterprise customization.
Bria’s commitment to the creator economy sets it apart. By partnering directly with content creators, the company ensures that those contributing to its training data are fairly compensated. This model not only amplifies artistic work but also creates value for intellectual property holders, including Hollywood’s premium content ecosystem. Bria’s approach redefines how AI companies engage with creative communities, prioritizing ethical innovation.
Bria’s open-source text-to-image model marks a significant advancement in responsible AI development. By combining high performance, ethical data practices, and enterprise-ready features, Bria empowers businesses and creators alike, driving innovation in a sustainable and inclusive manner.
Bria is the Visual Generative AI Platform-as-a-Service (PaaS) purpose-built for developers and builders to develop, deploy, and manage AI-powered visual solutions at scale. Its enterprise-grade platform delivers unprecedented speed and flexibility along with the safety, compliance, and quality needed for commercial production. Bria's foundation models, trained on 100% licensed data from 30+ partners, ensure legal compliance with proper attribution to data owners and artists.
TechIntelPro empowers B2B technology decision-makers with the latest industry trends, deep insights, leadership perspectives and actionable intelligence to drive both organizational and professional milestones—and thrive on the tech pulse.