airllm
AirLLM 70B inference with single 4GB GPU
- chinese-llm
- chinese-nlp
- finetune
- generative-ai
- instruct-gpt
- instruction-set
- llama
- llm
- lora
- open-models
- open-source
- open-source-models
- qlora
- Stars
- 35,906
- Forks
- 3,770
- + today
- +10
- Created
- 3y
Ranking data as of October 11, 2026 (UTC).
Star History
Today, hour by hour
Overview
AirLLM 70B inference with single 4GB GPU It ranks #41 on GitTiger, gaining +10 stars on October 11, 2026 (UTC).
The project is written in Jupyter Notebook and has 3,770 forks. It was created 3y ago.
What it is
AirLLM is a Jupyter Notebook project for running large language model inference with limited GPU memory. The description says it runs 70B-parameter models on a single 4 GB GPU.
Capabilities described
- The description claims 70B inference on a single 4 GB GPU. This is a promotional claim, and the supplied evidence does not include benchmarks. - The topics mention fine-tuning, LoRA, Chinese NLP and instruction-set models.
Getting started
The supplied evidence does not include setup steps or code samples. The README summary contains only a support link and a short project label, so see the repository README.
git clone https://github.com/lyogavin/airllm.git
cd airllm
# see README for setupFrequently Asked Questions
Can it run a 70B model on a 4 GB GPU?
The repository description makes that claim. The supplied evidence does not include benchmarks, so the claim is unverified here.
Does it support fine-tuning?
The topics include fine-tuning and LoRA, but the supplied evidence does not describe how to do it.
What license does it use?
The supplied metadata lists Apache-2.0.