Bria Artificial Intelligence
 | This article contains promotional content. (January 2026) (Learn how and when to remove this template message) |
Bria AI
Bria AI (known as Bria) is an Israeli- American artificial intelligence company that specializes in visual generative AI technology. The company develops enterprise-grade AI platforms for image generation, editing, and content creation, with a focus on copyright compliance and responsible AI practices.
History
Bria AI was founded in 2020 by Dr. Yair Adato.[1]The company is headquartered in Israel with operations in the UK and United States. In March 2025, Bria announced $40 million in Series B funding, bringing its total capital raised to $65 million.
Recognitions and Awards
In April 2025, Bria was named to CB Insights' AI 100 list, recognizing the 100 most innovative AI startups globally.[1] The company was also nominated for the TIME's 100 Most Influential Companies in 2026 [2] [4] and chosen as one of Forbes' Next Billion Dollar Startup 2025.[3] Bria was recognized as an AWS Design Partner of the Year in 2024 for its work in visual generative AI solutions.[4]
Technology
Platform Architecture
Bria operates a Platform-as-a-service (PaaS) model offering visual generative AI capabilities through multiple deployment options, including APIs, SDKs, and source code access.[5] The platform is built on cloud-native infrastructure using containerized Kubernetes architecture and operates across multiple regions on major cloud providers including Amazon Web Services, Microsoft Azure, and Google Cloud Platform.[5]
Fibo Model
In Q4 2025, Bria launched FIBO (also referred to as GAIA-Aplha), described as a "JSON-native text-to-image model".[6] The model is based on the DimFusion architecture and uses SmoILM3-3B as its language backbone, with the Wan 2.2 VAE operating at 1024x1024 latent solution.[6]
FIBO was trained on approximately 1 billion licensed image-caption pairs with structured JSON captions up to 1000+ words.[6] The model processes parameters including lighting, composition, color, camera settings, and depth of field, designed for controllable and reproducible outputs.[6]
According to Bria's published benchmarks, FIBO achieved an alignment score of 89.0 on the PRISM benchmark, outperforming open-source baselines including Qwen-Image (84.1) and HiDream-I1-Full (80.0).[6]
Licensed Training Data
Bria's models are trained exclusively on licensed data from over 30 content partners.[5] The company has established partnerships with major content providers including Getty Images for licensed training datasets.[5]
The company developed a patented attribution technology that tracks the contribution of training data to generated outputs, enabling compensation for data owners based on their content's influence on each generation.[16]
References
This article "Bria Artificial Intelligence" is from Wikipedia. The list of its authors can be seen in its historical and/or the page Edithistory:Bria Artificial Intelligence. Articles copied from Draft Namespace on Wikipedia could be seen on the Draft Namespace of Wikipedia and not main one.
- ↑ 1.0 1.1 "CB Insights AI 100 Announcement,". CB Insights. Retrieved April 2025. Check date values in:
|access-date=(help) Cite error: Invalid<ref>tag; name "CB Insights" defined multiple times with different content - ↑ "TIME 100 Most Influential Companies Nomination". TIME. Retrieved 2026. Check date values in:
|access-date=(help) - ↑ "Forbes Israel Next Billion Dollar Startup". Forbes Israel. Retrieved October 2025. Check date values in:
|access-date=(help) - ↑ "Announcing the 2024 Geo and Global AWS Partners of the Year". AWS Partners. Retrieved December 2024. Check date values in:
|access-date=(help) - ↑ 5.0 5.1 5.2 5.3 "Bria AI". Official Website. Retrieved December 2025. Check date values in:
|access-date=(help) Cite error: Invalid<ref>tag; name "Bria Website" defined multiple times with different content - ↑ 6.0 6.1 6.2 6.3 6.4 "FIBO on Hugging Face". Hugging Face. Retrieved November 2025. Check date values in:
|access-date=(help) Cite error: Invalid<ref>tag; name "Hugging Face" defined multiple times with different content
