# hyperionaidev/smolmaxxing

> A project for running LLMs on low-power devices (TV boxes, old tablets, SBCs) with ARM-specific optimizations: custom kernels, weight streaming, MoE expert paging. The goal is to squeeze maximum tokens per second out of devices with 1-2 GB of RAM.

- Magnitude: 6.8 out of 10 — Early signal
- Stars: 656 total · +23 stars measured 2026-11-14, 00:03–10:32 UTC, ≈ 115 by evening
- Star trust: star growth looks organic
- Category: Language models · Language: C · License: MIT · Created: 2026-11-04 · Last push: 2026-11-13
- GitHub: https://github.com/hyperionaidev/smolmaxxing · Homepage: https://hyperion-ai.com/products/smolmaxxing · Page: https://gitnova.dev/en/r/hyperionaidev/smolmaxxing

## Useful for

- Run an 8B MoE model on a 2 GB RAM TV box via expert paging from flash
- Speed up 1-bit model inference on an old iPad Air with a NEON lookup-table kernel
- Stream 9B model weights from flash on a device where the model doesn't fit in RAM

## Why it’s here

- Star-counter measurements on 2026-11-14 (UTC), 00:03–10:32: 633 → 656 stars (+23). This is the change over that interval.
- Estimated end-of-day forecast: about +115 stars, using observed gains and the previous day.
- The repository is 10 days old and already has 656 stars. With less than two weeks of history, there's no usual pace to compare the spike against yet.
- Hacker News: “I'm making LLMs work on small and old devices” — 475 points, 1 day ago.
- About 11 forks a day — people are taking the code.

## Star trust

Star growth looks organic.

Star-trust labels are heuristics based on the repository’s behavior, not a check of every stargazer.

## Numbers

- Forks: 23
- Issues and pull requests: 3
- Watchers: 2
- Average over the last week: 109 per day
- Usual pace: too little history (under two weeks)
- Stars in the last hour (measured): 2

## Stars per day, last 11 days (oldest → newest, today is partial)

2026-11-04 … 2026-11-14: 0, 0, 0, 0, 0, 0, 0, 1, 392, 257, 23

## Magnitude by day, last 3 days

11-12 6.6, 11-13 7.8, 11-14 7.2

## Hacker News

- I'm making LLMs work on small and old devices — 475 points, 180 comments: https://news.ycombinator.com/item?id=50468888

## Spotted in now

- Spotted on Hacker News

## Similar by description

1. **microsoft/BitNet** — 1.3 · Steady · Language models · C++ · +2 stars measured 2026-11-14, 00:08–10:38 UTC, ≈ 6 by evening
   Official inference framework for 1-bit LLMs (BitNet b1.58) with optimized kernels for CPU and GPU. Runs ternary models fast and losslessly, including a 100B model on a single CPU.
   Full card: https://gitnova.dev/en/r/microsoft/BitNet.md

---

Magnitude (0–10) measures how fast and how unusually interest in a repository is growing right now. It is not a quality score.
Days are UTC. “So far today” is a fact; “expected by the end of the day” is a forecast.
Summaries and use cases are written by an LLM (DeepSeek V4.1 Flash) from the README and may be inaccurate: verify specific claims (benchmarks, speed, hardware) in the repository itself.
Data as of 2026-11-14 10:41 UTC, updated every 30 minutes.