Meta’s open source AI model, Llama 3.2, is now deployed on the International Space Station (ISS) National Laboratory through a collaboration with Booz Allen, the advanced technology company delivering outcomes with speed for America’s most critical defense, civil, and national security priorities. This fine-tuned version of Llama 3.2 enables astronauts to use AI for research and operations in low Earth orbit. The goal of this effort isn't just to replace paper documents onboard the ISS National Lab, but also decrease reliance on instructions received from the ground while deployed at the edge which will result in lower costs, less compute power required, and faster outcomes when responding to onboard anomalies or maintenance issues, without depending on Earth-bound internet.

Why Llama for Space?

Llama’s open source design allows it to function without internet access, a key requirement for the ISS, where connectivity is unavailable. This offline capability ensures that sensitive data remains secure without needing to be sent to Earth or processed externally. Llama’s model weights numerical values that guide its decisions are publicly available, allowing researchers to adapt the AI to new conditions while keeping costs manageable. This flexibility suits the ISS National Laboratory, where scientists conduct experiments unique to microgravity.

The Technology Powering Space Llama

Space Llama operates using a combination of Booz Allen’s A2E2™ (AI for Edge Environments), Hewlett Packard Enterprise’s Spaceborne Computer-2, and NVIDIA’s accelerated computing. Optimized with NVIDIA CUDA software and libraries like cuDNN and cuBLAS, the system processes AI tasks in just over a second, compared to minutes previously. Its compact, energy-efficient design, similar to satellite systems, supports both generative AI, which creates content from prompts, and multimodal AI, which handles text, images, and audio. For instance, astronauts can access technical manuals or instructions quickly, without relying on Earth-based systems.

A Foundation of Prior Success

This project builds on Booz Allen’s earlier work, including the deployment of a generative AI model on the ISS in August 2024 via HPE’s Spaceborne Computer-2. In November 2024, Meta made Llama models available to U.S. government and private sector partners, enabling this collaboration. These steps highlight the practical use of open source AI in space.

As Booz Allen’s chief technology officer, Bill Vass, stated, “Space innovation has been limited historically due to reliance on Earth-based connectivity for compute and communications capabilities. Space Llama brings tools directly to the edge of space to quickly conduct critical repairs and maintain the ISS National Lab propelling us towards a future of space-based science, discovery, and the ability to operate at the farthest mission edge space.”

Enhancing ISS Research

Space Llama supports the ISS, the largest human-made structure in space, by providing AI tools for data analysis and operational tasks. It aids researchers in diverse scientific projects, from material science to biology, and sets the stage for future missions to the Moon, Mars, and beyond, as well as advancements in satellites, drones, and autonomous systems.

The Value of Open Source AI

Meta’s collaboration with Booz Allen demonstrates the importance of open source AI in strengthening U.S. capabilities in space and technology. Llama’s performance in the challenging environment of space shows its potential for broader applications on Earth. Open source models like Llama encourage collaboration, reduce costs, and support rapid adaptation.

A Step Toward the Future

Space Llama is an early milestone in integrating AI with space exploration. As open source models like Llama continue to develop, they may become essential tools for astronauts, aiding in complex problem-solving and scientific discovery. This project not only enhances research on the ISS but also signals the potential for AI in humanity’s exploration of space.