Meta's Llama language model is being tested as the brain behind military-grade augmented-reality glasses that could fundamentally change how soldiers interact with drones and battlefield information. The defense-tech company Anduril is developing smart glasses for the U.S. Army in partnership with Meta, using Llama alongside competing models from Google and Anthropic to translate soldiers' spoken commands into actionable military software instructions.

What Are These Military Smart Glasses Designed to Do?

Anduril is working on two separate augmented-reality headset projects. The first is the Army's Soldier Born Mission Command (SBMC) program, for which Anduril won a $159 million prototyping contract to develop glasses that attach to existing military helmets. The second is a self-funded project called EagleEye, which integrates the technology directly into a custom helmet design.

Both systems aim to overlay critical battlefield information onto a soldier's field of view, from simple compass directions to complex maps, drone locations, and AI-powered target recognition. The glasses would let soldiers issue commands in plain language, such as ordering an evacuation or planning a route that avoids restricted areas. Llama and other large language models (LLMs), which are AI systems trained on vast amounts of text to understand and generate human language, would translate these spoken instructions into commands the military software can execute.

How Would Soldiers Actually Control These Systems?

• Voice Commands: Soldiers speak to the interface in natural language, such as "send a drone to surveil that area," and the LLM converts the request into actionable instructions.
• Eye-Tracking Control: The system can track a soldier's eye movements to select targets or navigate menus without requiring hands or voice, useful in high-noise combat environments.
• Subtle Taps: Soldiers can communicate through tracked eye movements and subtle taps on the device, minimizing distraction and maintaining situational awareness.

Quay Barnett, the vice president leading Anduril's efforts and a former Army Special Operations Command officer, explained that his goal is to optimize "the human as a weapons system." The vision is for drones and soldiers to see together, share information seamlessly, and make decisions as one unified entity.

Why Is Meta's Llama Being Used in Military Applications?

Anduril is currently testing three different large language models: Google's Gemini, Meta's Llama, and Anthropic's Claude. The choice of multiple models reflects the company's effort to find the best fit for military operations, where accuracy and reliability are critical. Llama, Meta's open-weight AI model, offers flexibility and can be customized for specific military use cases. Open-weight models are AI systems whose underlying code and weights (the numerical parameters that define the model's behavior) are publicly available, allowing organizations to modify and deploy them on their own hardware.

The system is designed to handle multi-step tasks autonomously. For example, a soldier might instruct a drone to surveil an area and return once it identifies something resembling an artillery unit. The system would then recommend courses of action, such as sending another drone to strike the target, but these recommendations would still require approval from the normal chain of command.

What Technical Challenges Remain?

Multiple obstacles stand between prototype and battlefield deployment. Unlike Meta's commercial Ray-Ban smart glasses, these military versions must operate in environments filled with dust, explosions, and smoke. The computing power and battery life required to run advanced AI models locally on the device adds significant weight, a major concern for soldiers already carrying over 100 pounds of gear. Additionally, the glasses must function in areas without reliable 5G connectivity, meaning powerful computer vision and AI models need to run directly on the device rather than relying on cloud servers.

Jonathan Wong, a former U.S. Marine and senior policy researcher at RAND, emphasized the cognitive burden these systems impose. "How much mental bandwidth do you have to be both aware of your surroundings and to operate this technology in a way that makes you and your whole unit better?" he asked. Wong recalled his experience as a platoon commander managing a radio operating on three channels simultaneously, noting that when two people spoke at once on different channels, he lost comprehension of both messages and awareness of his surroundings. "I think there are limits to what you can take in," he observed.

"It's got to work, and it's got to be pretty seamless," Wong stated.

Jonathan Wong, Senior Policy Researcher at RAND

The Army is not expected to select a final design for production until 2028, if it chooses one at all. This timeline reflects lessons learned from Microsoft's failed $22 billion smart glasses contract, which was cancelled after a Pentagon audit revealed the Army had not properly tested the glasses before committing to massive spending.

What Role Does Meta's Hardware Play?

While Llama provides the AI language processing, Meta is also supplying much of the physical hardware for both Anduril projects, including the displays and waveguides that send visuals to the soldier's eye without blocking their view of the battlefield. This partnership marks a significant reconciliation between the two companies. In 2017, Facebook (now Meta) ousted Anduril founder Palmer Luckey following an internal conflict over his political views. The two organizations are now collaborating on military augmented-reality technology, while Meta's CEO Mark Zuckerberg has adopted a friendlier stance toward the current administration.

Anduril is also developing a new digital night vision system for both projects, using electronic sensors and algorithms to amplify low levels of light. This technology has been promised for decades but historically worked too slowly for practical use and produced grainy images. Anduril claims it has achieved improvements through techniques rooted in both generative AI and traditional machine learning.

What Happens If the Army Rejects These Glasses?

Even if the Army ultimately prefers a competing design over Anduril's EagleEye system, the company plans to market the glasses to foreign militaries. This dual-track approach gives Anduril a fallback market if the U.S. military procurement process favors another vendor. Anduril is not alone in this competition; Rivet, a company specializing in wearable military sensors, received a $195 million prototyping contract, and the Israeli defense-tech firm Elbit received a $120 million contract in March.

The integration of large language models like Meta's Llama into military hardware represents a significant escalation in how closely soldiers will rely on AI systems for decision-making. While computer vision models have long been used to identify objects in military applications, and chatbots have recently entered decision-making processes in some conflicts, these technologies have not yet reached most frontline soldiers. A smart glasses system tasked with identifying threats and recommending strikes introduces substantial new risks of AI errors in life-or-death situations.