In the Crest catalog
lyogavin/airllm
AirLLM 70B inference with single 4GB GPU
- Stars
- 35.3K
- Forks
- 3.7K
- Language
- Jupyter Notebook
- Last push
- 23d ago
- Last release
- 49d ago
Linked by the package’s own registry metadata.GitHub (opens in a new tab)
Package · PyPI
airllm
PyPI · v4.0.0 · Apache Software License
AirLLM runs 70B large language models on a single 4GB GPU without quantization, distillation or pruning. Kimi K3 2.8T on under 4GB, Qwen3.8-Flash-Next 125B on 6GB, DeepSeek-V3 671B on ~12GB. Train Qwen3.8-Flash-Next under 6GB.
- AI & Machine Learning (primary)
- LLM Apps & Clients
- LLM Infrastructure
- Open Source Infrastructure · Tentative
Tentative categories were assigned with moderate confidence (0.5 to 0.75): they are shown, but they do not count toward a category’s momentum. How categories work
Install
pip install airllmAdoption
- Adoption
- Not published by PyPI
- Not published
- Popularity
- Not published by PyPI
- Not published
- Dependents
- Not published by PyPI
- Not published
- Maintenance
- 0.86
- DerivedRelease cadence and recency, 0 to 1
- Releases per year
- 14
- 39 versions over its life
Growth by period
PyPI publishes no adoption figure, so there is no growth to derive. Nothing is estimated in its place.
- 7 days
- Not published by PyPI
- Not published
- 30 days
- Not published by PyPI
- Not published
- 90 days
- Not published by PyPI
- Not published
- 1 year
- Not published by PyPI
- Not published
Adoption history
Adoption history
PyPI publishes no adoption count, so there is no history to chart. That is a fact about the registry, not a gap in ours: the package’s releases and maintenance are measured instead.
Facts from the registry
- Latest version
- v4.0.0
- 39 versions
- First published
- NOV 17, 2023
- Last release
- SEP 5, 2026
- 49d ago
- Licence
- Apache Software License
Repository
Dependencies latest version · v4.0.0
Runtime 8
- accelerate
>=1.0 - huggingface-hubAny version
- safetensorsAny version
- scipyAny version
- sentencepieceAny version
- torch
>=2.4 - tqdmAny version
- transformers
<6,>=4.49
Related packages
- ttnnGeneral compute framework for Tenstorrent devicesSame primary category
- rapid-peRapidPE: The original low-latency gravitational wave parameter estimation code.Same primary category
- keras-hub-nightlyPretrained models for Keras.Same primary category
- amd-gaiaGAIA is a lightweight agent framework designed for the edge and AI PCs.Same primary category
- nemo-relayPython bindings for the NeMo Relay agent runtime.Same primary category
- pipecat-aiAn open source framework for voice (and multimodal) assistantsSame primary category
Related news
No published news event mentions this package or its repository yet.