Meta today announced the latest updates to its proprietary AI, Meta Llama 3.1. This release underscores Meta’s commitment to fostering innovation and collaboration within the AI community by providing developers, researchers, and businesses with new and improved cutting-edge AI tools.
Here are the key takeaways from today’s update briefing.
What’s New for Llama 3.1
Llama 3.1 now features updated 8B and 70B models, as well as the introduction of the groundbreaking Llama 3 405B – the industry’s largest and most advanced open model ever created, offering unmatched flexibility, control, and capabilities.
- Unparalleled Performance: Rivals the performance of all leading closed-source models.
- Expanded Language Support: Now covers 8 languages, including Thai.
- Enhanced Conversational Abilities: Expanded context length (128k) enables richer interactions.
- Stronger Reasoning: Tackles complex tasks with improved reasoning capabilities.
- Workflow Enhancements: Supports synthetic data generation and model distillation.
- Responsible AI: New system-level safety components ensure responsible use.
- Thriving Ecosystem: Over 30 partners are ready to support Llama 3.1 from day one.
Since the release of the Llama model’s first iteration last year, there have been over 300 million downloads and over 20,000 derivative models developed for a range of applications. Today, Meta AI is leading the pack both in flexibility and overall performance, and is more creative, intelligent, and accessible than ever before.
- Creativity: Users can unleash their imagination with the “Imagine Yourself” feature, generating AI images of themselves in any scenario. New AI editing tools also lets users fine tune images Meta AI generates – allowing users to add, remove or change objects as desired.
- Intelligence: Meta AI empowers users to tackle complex challenges with ease. Users benefit from more accurate answers to complex questions, especially on topics such as math and coding, and an evolving AI experience driven by continuous updates based on user feedback.
- Accessibility: Meta AI is being seamlessly integrated across Meta’s family of platforms and products such as WhatsApp, Facebook, and more. The addition of multiple new languages also brings Meta AI to even more users, with Llama’s open source model allowing developers to adapt and train their own models, protect their data, and build robust systems.
Llama 3.1 Thai Language Expansion
At today’s update briefing, Rafael Frankel, Meta’s Director of Public Policy for Southeast Asia, spoke on Meta’s plans for Llama 3.1 in Thailand: “Our latest release of Llama 3.1 is raising the bar again for openly available AI models and is the introduction of the industry’s largest and most advanced open model ever created – Llama 3 405B. The Llama 3.1 family also includes updated 8B and 70B models, which are now available in Thai language. This open source model will allow Thai developers to develop many more innovations for Thailand. We believe AI has more potential than any other modern technology to increase human productivity and creativity – and to accelerate economic growth while unlocking progress for businesses and communities around the world including Thailand.”
Developers can visit this site and access our Llama’s resources
Unlocking AI Opportunities for Businesses and Communities in Thailand
In addition to adding Thai as a supported Llama AI language, Meta will also be running multiple AI-focused programs and awarding grants worth a total of $2 million USD to AI developers and companies around the world, including in Thailand.
1. Meta’s Llama 3.1 Impact Grants
Meta’s Llama Impact Grants are a global call for proposals that utilize Llama AI and its features to empower social impact and address critical challenges in areas such as education, the environment, and open innovation.
Starting today, Meta is accepting applications for the Llama 3.1 Impact Grants. Eligible organizations are invited to submit proposals outlining how Llama 3.1 can be utilized to address social challenges in their communities, with selected winners receiving up to $500,000 USD in funding to support their projects. The announcement of the winners is expected in early 2025.
Learn more about the Llama 3.1 Impact Grants program and apply now to be considered.
2. APAC AI Accelerator Program
Meta will also host a series of regional programs to provide education, inspiration, and networking opportunities for AI developers in the APAC region.
One of these events is the APAC AI Accelerator Program, a regional competition where Llama developers will submit pitches that demonstrate how Llama AI can be used to solve critical social challenges in their country. The winners will then get to represent their country at a virtual Hackathon competition in Singapore for the chance to win up to $100,000 USD in specialized grants to support the development of their projects, with the most promising projects also being considered for the open call award of $500,000 USD.
In Thailand, Meta has partnered with the Digital Economy Promotion Agency (DEPA), the Electronic Transactions Development Agency (ETDA), the AI Governance Clinic and the Artificial Intelligence Association of Thailand (AIAT) to run the competition and nominate one winner to join the regional competition in Singapore.
3. AI Courses for SMEs in Thailand
Meta has also partnered with the Ministry of Commerce to offer AI marketing courses for Thai SMEs, equipping them with the skills and knowledge to leverage AI for growth.
In addition, Meta will host its Meta Marketing summit in late August, where over 800 SMEs will be invited to join the event and learn more about Meta’s AI tools and services.
Lack of secure gen-AI workloads a big challenge for Asia, says AWS expert
AIS with Oracle to Build Thailand’s First Locally Owned and Operated Hyperscale Cloud
