Llama 3.1 Impact Grants will incorporate both an open call for proposals as well as sourcing regionally specific use-cases from a series of events around the world. Winners will be identified in both tracks.

At least one event will be hosted in 5 regions before the end of October. These events will include hackathons, workshops, pitch competitions, and application-based demo days where Meta technical mentors will advise developers on how to optimize for building impactful applications with Llama 3.1.

Organizations participating in these global events will need to indicate their participation on their Llama 3.1 Impact Grant application and will be entered into an additional specialized track within the Grant program. While still eligible to receive up to $500,000 USD, at least one or more Organization per region will be selected from this track to receive up to $100,000 USD in order to continue iterating on their project. For those considering participating in these events and entering this specialized track, we encourage you to submit your application after attending the event.

• Showcase the newest capabilities of Llama 3.1 (e.g., multilinguality)
• Lower barriers for other developers to build with Llama (e.g., enhance features such as evaluation, data preparation, or prompt engineering)
• Address one of the following topics:
• • Economic development
  • • E.g., future of work; scaled business solutions; support for small businesses
  • Science and innovation
  • • E.g., climate, health, accessibility, agriculture; infrastructure support
  • Public services
  • • E.g., delivery of services; translation of government resources; solving challenges for policymakers; resource allocation; migration
  • Education and skills development
  • • E.g., improve learning outcomes; retraining; lifelong learning

The application window is now open!

The application will be open from August 5 - November 22, 2024. All proposals will be evaluated using a selection criteria that assesses:

• Demonstrated utility of Llama 3.1: Llama should be a fundamental component of the project or product, and its utility should be clearly defined. Preference will be given to those applicants who can demonstrate the utility of Llama 3.1’s new features, including multilinguality.
• Applicability: The proposed solution should measurably address social impact. Preference will be given to applicants who propose to address: economic development; science and innovation; public services; or education and skills development.
• Impact: The proposed solution should articulate a vision for how it impacts the broader problem space and provides social benefit. This might include how the solution works across geographies, addresses the needs of additional communities, or similar.
• Technical feasibility: Proposed solutions should evaluate the technical feasibility of the project, including the availability of computational resources, technical expertise, and time required for development and deployment; outline a robust plan for collecting, cleaning, and preparing the data; and detail the metrics and methods for evaluating the performance of the fine-tuned/ RAG model.
• Business plan: Applicants should demonstrate an understanding of the target market, including potential users and their needs. This includes presenting data, research, or other evidence of market demand and how the fine-tuned model provides a unique value proposition or solves a specific problem better than existing solutions.

Those teams whose proposals are selected as winners will be notified by April 2025. Applicants of interest and runners up may be notified by or beyond that time. Stay tuned for program updates.

We will also be hosting two webinars on October 2nd and 9th where you can ask additional questions prior to the submission deadline. For webinar details and registration information, please visit the links below.

• October 2 registration
• October 9 registration

Applications will open August 5, 2024 at 9:00am PST and are due by 11:59 pm PST on November 22, 2024.

You are not obliged to disclose the trained model weights, or the data employed in its training.

The number of awards will vary based on the caliber of applications received. At least one award per region will be granted to organizations who have participated in a qualifying Meta event in 2024, not to exceed $100,000 USD in total prize money per region. Grand prizes for open call applicants will also be awarded, not to exceed $500,000 USD per prize. Participants in qualifying events will be automatically considered eligible for awards under both tracks.

We will not accept files uploaded to the application. You can input text or url links to external materials, including links to news articles, blogs, or other public resources describing your use of Llama 3.1.

Yes! While technical expertise will be helpful to ensure your application contains sufficient evidence of your planned program or project, having an AI expert on staff is not required. Proposals will be evaluated using an equal weighted average between the evaluation categories listed on the program website.