Technology Innovation Institute Open-Sourced Falcon LLMs: A New AI Model That Uses Only 75 Percent of GPT-3’s Training Compute, 40 Percent of Chinchilla’s, and 80 Percent of PaLM-62B’s

Falcon-40B

Falcon-40B is a powerful decoder-only model developed by TII (Technology Innovation Institute) and trained on a vast amount of data consisting of 1,000B tokens from RefinedWeb and curated corpora. This model is available under the TII Falcon LLM License.

The Falcon-40B model is one of the best open-source models available. It surpasses other models such as LLaMA, StableLM, RedPajama, and MPT in performance, as demonstrated on the OpenLLM Leaderboard.

One of the notable features of Falcon-40B is its optimized architecture for inference. It incorporates FlashAttention, as introduced by Dao et al. in 2022, and multi-query, as described by Shazeer et al. in 2019. These architectural enhancements contribute to the model’s superior performance and efficiency during inference tasks.

It is important to note that Falcon-40B is a raw, pre-trained model, and further fine-tuning is typically recommended to tailor it to specific use cases. However, for applications involving generic instructions in a chat format, a more suitable alternative is Falcon-40B-Instruct.

Falcon-40B is made available under the TII Falcon LLM License, which permits commercial use of the model. Details regarding the license can be obtained separately.

A paper providing further details about Falcon-40B will be released soon. The availability of this high-quality open-source model presents a valuable resource for researchers, developers, and businesses in various domains.

Falcon 7B

Falcon-7B is a highly advanced causal decoder-only model TII (Technology Innovation Institute) developed. It boasts an impressive parameter count of 7B and has been trained on an extensive dataset of 1,500B tokens derived from RefinedWeb, further enhanced with curated corpora. This model is made accessible under the TII Falcon LLM License.

One of the primary reasons for choosing Falcon-7B is its exceptional performance compared to other similar open-source models like MPT-7B, StableLM, and RedPajama. The extensive training on the enriched RefinedWeb dataset contributes to its superior capabilities, as demonstrated on the OpenLLM Leaderboard.

Falcon-7B incorporates an architecture explicitly optimized for inference tasks. The model benefits from integrating FlashAttention, a technique introduced by Dao et al. in 2022, and multi-query, as described by Shazeer et al. in 2019. These architectural advancements enhance the model’s efficiency and effectiveness during inference operations.

It is worth noting that Falcon-7B is available under the TII Falcon LLM License, which grants permission for commercial utilization of the model. 

Detailed information about the license can be obtained separately.

While a paper providing comprehensive insights into Falcon-7B is yet to be published, the model’s exceptional features and performance make it an invaluable asset for researchers, developers, and businesses across various domains.


Check out the Resource Page, 40-B Model, and 7-B Model. Don’t forget to join our 22k+ ML SubRedditDiscord Channel, and Email Newsletter, where we share the latest AI research news, cool AI projects, and more. If you have any questions regarding the above article or if we missed anything, feel free to email us at Asif@marktechpost.com

🚀 Check Out 100’s AI Tools in AI Tools Club

Niharika is a Technical consulting intern at Marktechpost. She is a third year undergraduate, currently pursuing her B.Tech from Indian Institute of Technology(IIT), Kharagpur. She is a highly enthusiastic individual with a keen interest in Machine learning, Data science and AI and an avid reader of the latest developments in these fields.

✅ [Featured Tool] Check out Taipy Enterprise Edition