Mistral AI Models Run Aboard First Satellite of 50-Satellite AI Constellation

Mistral’s AI models now run aboard a satellite in orbit, as Orbitworks and Loft Orbital launched the first spacecraft in a planned 50-satellite AI constellation. The satellite carries NVIDIA Jetson AGX Orin modules to run AI models onboard, enabling it to analyse data without sending raw imagery to Earth first.

The satellite forms part of a $1 billion programme announced in Paris in September. Orbitworks, which built the spacecraft at its Abu Dhabi facility, is developing the constellation with Loft Orbital. The companies announced the satellite’s capabilities, including onboard AI processing and support for third-party applications.

Mistral’s Ministral 3 models, in 3-billion- and 8-billion-parameter versions, run on the satellite. According to Mistral’s LinkedIn post, the models summarise acquisition data and answer operator queries directly in orbit. 

The spacecraft also runs nearly 20 other AI models for tasks such as ship detection, wildfire monitoring, flood mapping and image enhancement.

The constellation aims to reduce the time between capturing satellite imagery and delivering actionable information. Instead of transmitting raw images for ground processing, the satellites analyse observations in orbit and transmit detected events. 

This could allow users to receive alerts within seconds rather than waiting hours for imagery files. The companies are targeting applications in maritime monitoring, wildfire and disaster response, and the security of ports and critical infrastructure. 

Users can select applications from multiple providers through an AI app store. These applications will run on shared satellite infrastructure, allowing one constellation to serve several programmes.

The first spacecraft carries two NVIDIA Jetson AGX Orin modules. Its onboard systems run computer vision and vision-language models using NVIDIA CUDA 12.6, TensorRT and JetPack 6.2.1. The GPU also corrects images against known ground reference points, helping locate detections accurately on maps. The satellite encrypts onboard data using ML-KEM-1024, a post-quantum cryptographic standard.

The satellite also carries models developed by organisations including NASA’s Jet Propulsion Laboratory, Google, IBM, Meta, Microsoft and Hugging Face. These models support tasks ranging from cloud masking and scene description to detecting ships and mapping burned areas.

“NVIDIA compute sits at the centre of how we design our satellites,” said Paul Lasserre, General Manager of AI for Space at Loft Orbital. He said the technology helps convert satellite observations into real-time alerts and supports increasingly autonomous constellation operations.

Loft Orbital has operated AI-equipped satellites since March 2024, when it launched YAM-6 with an NVIDIA GPU. The company says it has since flown more than 20 AI missions. Earlier in 2026, it also demonstrated an onboard vision-language model with NASA’s Jet Propulsion Laboratory using its YAM-9 satellite.

The wider programme will expand the constellation to 50 satellites, all equipped with NVIDIA accelerated computing. The infrastructure will serve government customers in France, Europe and the UAE, alongside global commercial and sovereign clients.

Loft Orbital is evaluating NVIDIA’s Space-1 Vera Rubin compute module and IGX Thor for future satellites. These systems could support more advanced AI reasoning and real-time processing as the constellation expands.