--- title: NASA Puts Google’s Gemma Large Language Model in Orbit description: NASA's Jet Propulsion Laboratory has demonstrated the first in-orbit use of a vision-language model, deploying Google's Gemma 3 4B on a Loft Orbital YAM-9... canonical: https://daily.dev/posts/nasa-puts-google-s-gemma-large-language-model-in-orbit-qojsyvtlx twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: NASA Puts Google’s Gemma Large Language Model in Orbit | daily.dev og:description: NASA's Jet Propulsion Laboratory has demonstrated the first in-orbit use of a vision-language model, deploying Google's Gemma 3 4B on a Loft Orbital YAM-9... og:url: https://daily.dev/posts/nasa-puts-google-s-gemma-large-language-model-in-orbit-qojsyvtlx og:image: https://api.daily.dev/og/posts/QojsyvtlX.png og:image:alt: NASA Puts Google’s Gemma Large Language Model in Orbit og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # NASA Puts Google’s Gemma Large Language Model in Orbit **[IEEE Spectrum](https://daily.dev/sources/ieeespectrum)** · 6 min read · 1 upvotes · 0 comments ## Summary NASA's Jet Propulsion Laboratory has demonstrated the first in-orbit use of a vision-language model, deploying Google's Gemma 3 4B on a Loft Orbital YAM-9 satellite. The NAVI-Orbital system uses a LangGraph-based agentic framework and a 4-bit quantized Gemma 3 model running on an Nvidia Jetson Orin AGX to analyze satellite imagery directly in space. It achieved 88% accuracy on an image classification benchmark without any fine-tuning. Two live capture tests were conducted over Toulouse, France and the coast of Argentina. The key benefit is 'semantic compression': instead of downlinking large raw image files over limited bandwidth, the satellite sends concise text descriptions, potentially cutting wildfire detection delays from 90 minutes to near-real-time. The system also signals a new interaction paradigm where scientists can control spacecraft behavior via natural language prompts rather than structured commands. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### What model and hardware did NASA use to run an LLM on a satellite in orbit? NASA's Jet Propulsion Laboratory ran a 4-bit quantized version of Google's Gemma 3 4B model on an Nvidia Jetson Orin AGX aboard Loft Orbital's YAM-9 satellite. The quantized model requires only 8GB of memory, letting it run on the low-power compute module. The system, called NAVI-Orbital, coordinates operations through a LangGraph-based conductor and produces plain-text descriptions of satellite images. _Developers exploring edge deployment of quantized LLMs can follow real-world benchmarks like this one on daily.dev._ ### How accurate was Gemma 3 at classifying satellite images without fine-tuning? Gemma 3 4B achieved 88 percent accuracy classifying images in a ground-based benchmark of 7,960 images, using the same base model available for download on Hugging Face, with no training or fine-tuning on that dataset or its categories. The model was later tested live in orbit over Toulouse, France, and the coast of Argentina. _Anyone weighing zero-shot vision-language models for real deployments can track results like this on daily.dev._ ### Why would a satellite use an AI model to describe images instead of sending raw imagery to the ground? Using an AI model onboard lets a satellite downlink a short text summary instead of large raw image files, a technique called semantic compression, since bandwidth to ground stations is often limited and only available at set intervals. This can cut delays in time-sensitive applications like wildfire detection, which currently can be delayed by up to 90 minutes due to bandwidth and processing limits. _Engineers designing bandwidth-constrained AI pipelines can follow approaches like semantic compression on daily.dev._ ## Similar posts on daily.dev - [Google’s Gemma 4 shines on local systems – both big and small](https://daily.dev/posts/google-s-gemma-4-shines-on-local-systems-both-big-and-small-qpert58dk) · InfoWorld · 0 upvotes · 0 comments - [Gemma 4 VLA Demo on Jetson Orin Nano Super](https://daily.dev/posts/gemma-4-vla-demo-on-jetson-orin-nano-super-jz1knl29n) · Hugging Face · 0 upvotes · 0 comments - [Google Updates Gemma 4: The best Small LLM is back \!\!](https://daily.dev/posts/google-updates-gemma-4-the-best-small-llm-is-back--tvolamyq7) · Medium · 1 upvotes · 0 comments - [Google's free Gemma 4 model runs on hardware you probably already own](https://daily.dev/posts/google-s-free-gemma-4-model-runs-on-hardware-you-probably-already-own-u5po7zub9) · XDA Developers · 6 upvotes · 0 comments --- Tags: [#llm](https://daily.dev/tags/llm), [#nasa](https://daily.dev/tags/nasa), [#local-ai](https://daily.dev/tags/local-ai), [#gemma](https://daily.dev/tags/gemma) [View this post on daily.dev](https://daily.dev/posts/nasa-puts-google-s-gemma-large-language-model-in-orbit-qojsyvtlx) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"NASA Puts Google’s Gemma Large Language Model in Orbit","url":"https://daily.dev/posts/nasa-puts-google-s-gemma-large-language-model-in-orbit-qojsyvtlx","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/nasa-puts-google-s-gemma-large-language-model-in-orbit-qojsyvtlx"},"datePublished":"2026-07-23T13:08:23.977Z","dateModified":"2026-09-14T08:42:00.749Z","description":"NASA's Jet Propulsion Laboratory has demonstrated the first in-orbit use of a vision-language model, deploying Google's Gemma 3 4B on a Loft Orbital YAM-9...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/458028df2c9ded44a1170bb868eeaa84?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/458028df2c9ded44a1170bb868eeaa84?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"IEEE Spectrum","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"IEEE Spectrum","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/b123f5e8420e492ab36f6adb30c1793a","url":"https://daily.dev/sources/ieeespectrum"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/nasa-puts-google-s-gemma-large-language-model-in-orbit-qojsyvtlx","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":1},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"llm,nasa,local-ai,gemma","timeRequired":"PT6M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"IEEE Spectrum","item":"https://daily.dev/sources/ieeespectrum"},{"@type":"ListItem","position":3,"name":"NASA Puts Google’s Gemma Large Language Model in Orbit"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/nasa-puts-google-s-gemma-large-language-model-in-orbit-qojsyvtlx#faq","mainEntity":[{"@type":"Question","name":"What model and hardware did NASA use to run an LLM on a satellite in orbit?","acceptedAnswer":{"@type":"Answer","text":"NASA's Jet Propulsion Laboratory ran a 4-bit quantized version of Google's Gemma 3 4B model on an Nvidia Jetson Orin AGX aboard Loft Orbital's YAM-9 satellite. The quantized model requires only 8GB of memory, letting it run on the low-power compute module. The system, called NAVI-Orbital, coordinates operations through a LangGraph-based conductor and produces plain-text descriptions of satellite images. Developers exploring edge deployment of quantized LLMs can follow real-world benchmarks like this one on daily.dev."}},{"@type":"Question","name":"How accurate was Gemma 3 at classifying satellite images without fine-tuning?","acceptedAnswer":{"@type":"Answer","text":"Gemma 3 4B achieved 88 percent accuracy classifying images in a ground-based benchmark of 7,960 images, using the same base model available for download on Hugging Face, with no training or fine-tuning on that dataset or its categories. The model was later tested live in orbit over Toulouse, France, and the coast of Argentina. Anyone weighing zero-shot vision-language models for real deployments can track results like this on daily.dev."}},{"@type":"Question","name":"Why would a satellite use an AI model to describe images instead of sending raw imagery to the ground?","acceptedAnswer":{"@type":"Answer","text":"Using an AI model onboard lets a satellite downlink a short text summary instead of large raw image files, a technique called semantic compression, since bandwidth to ground stations is often limited and only available at set intervals. This can cut delays in time-sensitive applications like wildfire detection, which currently can be delayed by up to 90 minutes due to bandwidth and processing limits. Engineers designing bandwidth-constrained AI pipelines can follow approaches like semantic compression on daily.dev."}}]} ```