TII Releases Falcon 2 AI Model, Outperforms Meta's Llama 3

Abu Dhabi-based The Technology Innovation Institute (TII) has launched Falcon 2, a series of AI models that includes Falcon 2 11B and Falcon 2 11B VLM. These models surpass existing benchmarks. 

Falcon 2 11B is a text-based model, while Falcon 2 11B VLM is a vision-to-language model that can convert visual inputs into textual outputs. Both models are multilingual and open-source, providing developers worldwide with unrestricted access.

Falcon 2 11B has been verified to outperform Meta’s Llama 3, which has 8 billion parameters, and performs on par with Google's Gemma 7B model, according to evaluations by Hugging Face Leaderboard. 

These models are designed to run efficiently on a single GPU, making them scalable and easy to integrate into various infrastructures, from high-end servers to personal laptops.

In a move to further enhance Falcon 2’s capabilities, TII plans to incorporate 'Mixture of Experts' (MoE). This advanced machine learning technique involves combining specialised smaller networks to improve performance by delivering more accurate and faster decision-making.

The models are released under the TII Falcon License 2.0, a permissive Apache 2.0-based software license promoting responsible AI use. More information is available at FalconLLM.TII.ae.

H.E. Faisal Al Bannai, secretary general of ATRC and Strategic Research and Advanced Technology Affairs Advisor to the UAE President, commented on the launch, saying, “While Falcon 2 11B has demonstrated outstanding performance, we reaffirm our commitment to the open-source movement and the Falcon Foundation.”

The first Falcon model, released in 2022, established TII's commitment to open-source AI, focusing on large language models (LLMs). It was trained on 1 trillion tokens and featured 7 billion parameters. This model builds on this foundation, enhancing capabilities.