I'm making LLMs work on small and old devices
What happened
A developer has created a project called smolmaxxing to run large language models on small, old devices like TV boxes and routers. The system uses custom kernels and memory techniques to keep only part of a model in RAM while streaming the rest from flash storage. This allows 8B and 16B mixture-of-experts models to run on devices with only about 1 GB of RAM. The project includes specific optimizations for different chip architectures, such as Cortex-A55 and A7, which lack certain instructions. All performance numbers were measured directly on the hardware, and the code is available under an MIT license.
Why it matters
This makes AI accessible on hardware you likely already own, such as old tablets or smart TVs, without needing expensive new computers. It demonstrates that powerful language models can function on minimal resources, potentially extending the life of older electronics.
Reality check
An automated, sceptical read of the sourceNotable. Running 16B parameter models on 1 GB of RAM is a significant technical achievement for low-resource environments, though it relies on specific hardware constraints and niche use cases.
Is this a big deal?
Get stories like this every morning
The day's AI news in plain English, in your inbox at 8:30 IST. Free.
More AI news
AI unlocks Gödel's notebooks, said to contain hundreds of proved theoremsIndustry news · 13 Nov 2026Gemini 4 Hydrogen and OrchestratorIndustry news · 12 Nov 2026Gemini 4 HydrogenIndustry news · 12 Nov 2026United Airlines bans AI agents from booking flights on behalf of travelersIndustry news · 11 Nov 2026My AI agents found an unmapped Maya site?Industry news · 11 Nov 2026This page was collected from github.com and summarised automatically by AIPulsee's pipeline. See how it runs.