New Model Ranks #1 on Hugging Face Leaderboard for Open Access LLMs
Model with 180 Billion Parameters is Trained on 3.5 Trillion Tokens, with 4 times the Compute Resources of Meta’s LLaMA 2
Falcon 180B is Open Access for Researchers and Commercial Users
Abu Dhabi-UAE: 06 September, 2023 – The Technology Innovation Institute (TII) in the United Arab Emirates (UAE) is pushing the boundaries of generative AI once again with the launch of Falcon 180B, an advanced iteration of its flagship large language model (LLM). This groundbreaking release strengthens the UAE's dominance in AI, offering Falcon 180B as an open access model for research and commercial purposes.
Following the remarkable success of Falcon 40B, an open source AI model that swiftly ascended to the top of the Hugging Face Leaderboard for LLMs in May 2023, TII, the applied research pillar of Abu Dhabi’s Advanced Technology Research Council (ATRC) continues to lead the charge in generative AI. Falcon 40B marked one of the first instances of open source models for both researchers and commercial users, and it was considered a pioneering leap in the field.
H.E. Faisal Al Bannai, Secretary General of the Advanced Technology Research Council, emphasized the positive impact of Falcon on the AI landscape and said: “We envision a future where the transformative power of AI is within everyone's reach. We are committed to democratizing access to advanced AI, as our privacy and the potential impact of AI on humanity should not be controlled by a select few. While we may not have all the answers, our resolve remains unwavering: to collaborate and contribute to the open source community, ensuring that the benefits of AI are shared by all.”
With a staggering 180 billion parameters and trained on 3.5 trillion tokens, Falcon 180B soars to the top of the Hugging Face Leaderboard for pretrained LLMs. It outperforms notable competitors like Meta’s LLaMA 2 in various benchmarks, including reasoning, coding, proficiency, and knowledge tests.
Among the best closed source LLMs, Falcon 180B ranks just behind OpenAI's latest GPT 4 and is on par with the performance of Google's PaLM 2 Large, the model powering Bard – despite being half the size of the model. The licensing framework for the model is established on ‘Falcon 180B TII License’, which is based upon Apache 2.0.
Dr. Ebtesam Almazrouei, Executive Director and Acting Chief Researcher of the AI Cross-Center Unit at TII, said: "The launch of Falcon 180B exemplifies our dedication to advancing the frontiers of AI, and we are thrilled to share its limitless potential with the world. Falcon 180B heralds a new era of generative AI, where the potential of scientific advancement is made available through open access to fuel the innovations of tomorrow. As we delve into frontiers of science and technology, our vision extends far beyond innovation; it's about nurturing a profound connection to address global challenges through collaborative breakthroughs.”
With over 12 million developers adopting and deploying the first release of Falcon, this significant upgrade is poised to become the premier model for various domains, from chatbots to code generation, and beyond.
Falcon 180B is compatible with the following major languages: English, German, Spanish, and French, with limited capabilities in Italian, Portuguese, Polish, Dutch, Romanian, Czech, and Swedish.