# airockchip/rknn-llm

Repository: https://github.com/airockchip/rknn-llm
Canonical: https://ross.abutalabs.com/products/rknn-llm
Language: Python
License: NOASSERTION
License Family: other
Last push: 2026-06-17T09:39:19+00:00

## Health v2 (maintenance only)
Score: 79/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 88, release rhythm 77, longevity 64
- inputs: {"age_days": 902, "days_push": 77, "days_rel": 77, "gap_med": 66.0, "n_releases_24m": 9}
- flags: no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1644, forks 219 (observed 2026-08-28T04:05:15.996678+00:00)

## What it is
RKLLM is Rockchip's software stack for converting, quantizing, and running large language models on Rockchip NPU chips such as RK3588 and RK3576. It includes a Python toolkit for PC-side model conversion and a C/C++ runtime for on-device inference.

## Use cases
- run an LLM locally on an RK3588 board
- convert a HuggingFace model like Qwen or Llama to RKLLM format
- quantize large language models for Rockchip NPU inference
- build a multimodal vision-language demo on embedded hardware
- deploy chat models offline on edge devices with no cloud connection

## When to choose
- you are deploying LLMs on Rockchip RK3588/RK3576/RK3562 hardware
- you need fully local, offline inference on embedded NPU devices
- you want to convert and quantize popular open models (Qwen, Llama, Gemma, Phi) for NPU acceleration

## When to avoid
- you are targeting GPUs, CPUs, or non-Rockchip accelerators
- you need server-scale LLM serving with high throughput
- your hardware lacks a Rockchip NPU

## Facets
- artifact type: library
- maturity: active
- function: llm-inference, machine-learning, sdk, serialization
- domain: large-language-models, machine-learning, embedded-systems, hardware
- platform: embedded, python, cpp
- tags: rockchip, npu, rk3588, model-conversion, quantization, edge-ai, on-device-inference, linux

## Member repositories
- airockchip/rknn-llm (main) score 79

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:05:15.996678+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T03:45:48.441125+00:00, confidence not recorded.
  - readme: https://github.com/airockchip/rknn-llm (fetched 2026-08-28T04:05:15.996678+00:00, sha d61583bd73eb)
- Data as of 2026-08-30T08:39:29.467469+00:00.