The UAE’s Falcon 3 challenges open source leaders amid the growing demand for small artificial intelligence models

UAE government supported Institute of Technological Innovation TII has announced the launch of Falcon 3, a family of open source small language models (SLMs) designed to run efficiently on lightweight GPU-based infrastructures.

Falcon 3 features four model sizes – 1B, 3B, 7B and 10B – with basic and education variants, promising to democratize access to advanced AI capabilities for developers, researchers and businesses. According to the Hugging Face leaderboard, the models are already outperforming or closely matching their popular open source counterparts in the size class, including Meta’s Llama and the category leader Qwen-2.5.

Development comes at a time Demand for SLMs,With fewer parameters and simpler designs than LLMs,,they are growing rapidly due to their efficiency, affordability,,and ability to deploy on devices with limited resources. It is suitable for a range of applications across industries, such as customer service, healthcare, mobile applications and the Internet of Things, where typical LLM programs may be too computationally expensive to run effectively. according to Evaluates reportsThe market for these models is expected to grow at a compound annual growth rate of approximately 18% over the next five years.

What does Falcon 3 bring to the table?

The Falcon 3 family is trained on 14 trillion tokens – more than double its Falcon 2 predecessor – and uses a decoder-only architecture with attention to batch querying to share parameters and reduce memory usage for key-value (KV) caching during inference. This enables faster and more efficient operations when dealing with various text-based tasks.

Basically, Forms supports four primary languages ​​– English, French, Spanish and Portuguese – and comes with a 32KB context window, allowing it to handle long inputs, such as strongly accented documents.

“Falcon 3 is versatile, designed for both general purpose and specialized tasks, offering tremendous flexibility to users. Its basic model is ideal for generative applications, while its instruction variant excels in conversational tasks such as customer service or virtual assistants.” Website.

According to Leaderboards On Hugging Face, while all four Falcon 3 models perform fairly well, the 10B and 7B are the stars of the show, achieving state-of-the-art results in reasoning, language comprehension, following instructions, code tasks, and math.

Among models within the 13B parameter size category, Falcon 3 10B and 7B versions outperform competitors, including Google Gemma 2-9BMeta Lama 3.1-8b, Mistral-7B,Wei 1.5-9B. It even outperforms Alibaba’s category-leading Qwen 2.5-7B in most benchmarks — such as MUSR, MATH, GPQA, and IFEval — except for MMLU, a test to assess how well linguistic models understand and process human language.

Publishing across industries

With Falcon 3 models now available on Face hugging,TII aims to serve a wide range of users, enabling cost-effective ,AI deployments without computational bottlenecks. With their ability to handle specific, domain-focused tasks with fast processing times, the models can run numerous applications at the edge and in privacy-sensitive environments, including customer service chatbots, personalized recommendation systems, data analysis, fraud detection, healthcare diagnostics, Supply chain improvement and education.

The institute also plans to expand the Falcon family further by introducing models with multimedia capabilities. These models are expected to be launched sometime in January 2025.

It is worth noting that all models have been released under the TII Falcon License 2.0, an Apache 2.0-based license with an acceptable use policy that encourages the development and deployment of responsible AI. To help users get started, TII has also launched Falcon Playground, a testing environment where researchers and developers can try out Falcon 3 models before integrating them into their applications.

Daily insights into business use cases with VB Daily

If you want to impress your boss, VB Daily has you covered. We give you the inside scoop on what companies are doing with generative AI, from organizational transformations to hands-on deployments, so you can share insights to maximize ROI.

Read our privacy policy

Thanks for subscribing. Check more VB Newsletters are here.

An error occurred.