About Falcon LLM
Falcon LLM is the family of generative AI models built by the Technology Innovation Institute in Abu Dhabi, and this site is where the family is published, documented and demonstrated. Each release gets an overview page with technical notes, and several models are exposed through hosted demos that can be tried in the browser before anything is downloaded.
The range has widened well past plain text generation. A hybrid architecture family combines attention based and state space modeling so that long responses and large batch sizes stay efficient, and a reasoning focused seven billion parameter variant targets mathematics and coding. Arabic is a clear priority: there are models tuned for Modern Standard Arabic and its dialects, a seven billion parameter model built around Emirati vocabulary, expressions and cultural context, a speech recognition model for Emirati and Arabic audio, and a lightweight model for extracting text from Arabic images. A multimodal perception model combines vision and language in one architecture so a system can read and interpret images from natural language prompts. Earlier generations, including an eleven billion parameter model and its vision language counterpart, remain listed.
The site publishes no prices, plans or subscription tiers, and the usual pricing paths return nothing. It operates as a model release and research hub, with models made available for download and trial rather than sold, which puts the running cost on whatever infrastructure or hosting a team chooses. The institute frames the work as making advanced AI available broadly rather than as a commercial product line.