Sticking to its plan, tech giant Meta just unveiled its latest Llama 3.1 Family of Models which entails a total of three different variations. This includes Llama 3.1 8B, 3.1 70B, and 3.1 405B. All these variants come with enhanced 128K contextual length.
Facebook’s parent firm will also now give developers the chance to make use of the outputs arising from the latest lineup to enhance its other models.
The news comes as the company also revealed the benchmarks that are said to be a great and competitive offering to other similar models by archrivals in the industry.
Furthermore, Meta evaluated the performance of the new models over more than 150 benchmark datasets which is a lot. They even carried out a host of human reviews to see the differences between the models and others on offer in the tech world today.
As per the results, we can confirm that the 405B is making heads turn and yesterday, it was said to outperform world leader GPT-4o in several aspects too. This was when data was leaked through Reddit. It similarly outperformed Claude and Gemini and there was said to be a force to be reckoned with in the industry.
Now, Meta has gone as far as to confirm how the new design is not only the biggest model but also an open-source foundation model. Meanwhile, other smaller Llama models are also available and can be found in both open and closed designs of similar sizes.
The latest Llama 3.1 family of variants is used by developers through the likes of AWS, Dell, Google Cloud, NVIDIA, Grog, and even Databricks. Meanwhile, the Llama 3.1 405B is found through Azure’s severless API endpoint too. You can even find Meta’s latest Llama variants on the Catalog for Azure AI Model.
So many tech giants today are working to design closed models, Meta mentioned. But it appears that the future is Open Source model designs which can close the gap. In 2023 alone, Llama 2 was said to be limited in comparison to just older gen variants behind Frontier. But in 2024, it’s obvious that things are changing at a quick pace.
Llama 3 is paving the way with not only the most advanced systems but also showing great leadership in certain locations. Beginning in 2025, the future Llama models tend to become some of the most advanced in this industry. And even before that can be confirmed, it’s great due to its cost-effectiveness, flexibility, and openness.
The news release of Meta’s Llama 3.1 is giving rise to a serious milestone in today’s leading open-source AI world. The performance is rivaling when you compare this to closed models and provides a serious commitment in terms of accessibility.
In this way, Meta hopes to make more AI capabilities possible while giving rise to more innovation putting out the latest standard for greater transparency and fostering new partnerships.
Zuckerberg was all praise for the latest launch and confirmed that the model is up for downloads. The company has also updated licensing to enable training out the model outputs.
It’s clear that Meta’s CEO Mark Zuckerberg envisions a future where open AI can really help developers and this new launch of Llama can transform to be the Linux of AI. He confirmed that the new launch of Llama already outweighs the performance and features offered by other advanced models of the industry and therefore there’s no reason why Meta’s future Llama cannot lead the tech world in terms of advancement.
Read next: Alphabet Reports Q2 2024 Gains: Google’s Revenue Up 14%, YouTube Revenue Rises to $8.67B
Facebook’s parent firm will also now give developers the chance to make use of the outputs arising from the latest lineup to enhance its other models.
The news comes as the company also revealed the benchmarks that are said to be a great and competitive offering to other similar models by archrivals in the industry.
Furthermore, Meta evaluated the performance of the new models over more than 150 benchmark datasets which is a lot. They even carried out a host of human reviews to see the differences between the models and others on offer in the tech world today.
As per the results, we can confirm that the 405B is making heads turn and yesterday, it was said to outperform world leader GPT-4o in several aspects too. This was when data was leaked through Reddit. It similarly outperformed Claude and Gemini and there was said to be a force to be reckoned with in the industry.
Now, Meta has gone as far as to confirm how the new design is not only the biggest model but also an open-source foundation model. Meanwhile, other smaller Llama models are also available and can be found in both open and closed designs of similar sizes.
The latest Llama 3.1 family of variants is used by developers through the likes of AWS, Dell, Google Cloud, NVIDIA, Grog, and even Databricks. Meanwhile, the Llama 3.1 405B is found through Azure’s severless API endpoint too. You can even find Meta’s latest Llama variants on the Catalog for Azure AI Model.
So many tech giants today are working to design closed models, Meta mentioned. But it appears that the future is Open Source model designs which can close the gap. In 2023 alone, Llama 2 was said to be limited in comparison to just older gen variants behind Frontier. But in 2024, it’s obvious that things are changing at a quick pace.
Llama 3 is paving the way with not only the most advanced systems but also showing great leadership in certain locations. Beginning in 2025, the future Llama models tend to become some of the most advanced in this industry. And even before that can be confirmed, it’s great due to its cost-effectiveness, flexibility, and openness.
The news release of Meta’s Llama 3.1 is giving rise to a serious milestone in today’s leading open-source AI world. The performance is rivaling when you compare this to closed models and provides a serious commitment in terms of accessibility.
In this way, Meta hopes to make more AI capabilities possible while giving rise to more innovation putting out the latest standard for greater transparency and fostering new partnerships.
Zuckerberg was all praise for the latest launch and confirmed that the model is up for downloads. The company has also updated licensing to enable training out the model outputs.
It’s clear that Meta’s CEO Mark Zuckerberg envisions a future where open AI can really help developers and this new launch of Llama can transform to be the Linux of AI. He confirmed that the new launch of Llama already outweighs the performance and features offered by other advanced models of the industry and therefore there’s no reason why Meta’s future Llama cannot lead the tech world in terms of advancement.
Read next: Alphabet Reports Q2 2024 Gains: Google’s Revenue Up 14%, YouTube Revenue Rises to $8.67B
