librariesAI / ML
AI / ML libraries
64 guides in this shelf, each one written to answer whether you should use the thing, not to sell it.
What our runs found on this shelf
We installed 64 of these 64 packages in a clean container and recorded what happened. The median install took 1.4s and left 64 MB on disk across 18 packages. sentencepiece was the quickest at 0.2s. The lightest was pgvector at 1 MB; the heaviest, sglang, pulled 10112 MB. 5 of them reported at least one known vulnerability at install time.
Numbers come from our own runs, not from the packages\u2019 own claims. How we test \u2192
- openai · PyPI
- huggingface-hub · PyPI
- @modelcontextprotocol/sdk · npm
- scikit-learn · PyPI
- langchain · PyPI
- mcp · PyPI
- tiktoken · PyPI
- openai · npm
- @anthropic-ai/sdk · npm
- tokenizers · PyPI
- langchain-core · PyPI
- anthropic · PyPI
- google-genai · PyPI
- ai · npm
- transformers · PyPI
- datasets · PyPI
- langsmith · PyPI
- litellm · PyPI
- safetensors · PyPI
- torch · PyPI
- onnxruntime · PyPI
- @ai-sdk/openai · npm
- fastmcp · PyPI
- nltk · PyPI
- langgraph · PyPI
- opencv-python · PyPI
- strands-agents · PyPI
- xgboost · PyPI
- langchain-community · PyPI
- pgvector · PyPI
- sentencepiece · PyPI
- claude-agent-sdk · PyPI
- mlflow · PyPI
- @langchain/core · npm
- sentence-transformers · PyPI
- spacy · PyPI
- scikit-image · PyPI
- einops · PyPI
- thinc · PyPI
- genai-prices · PyPI
- langfuse · PyPI
- mistralai · PyPI
- lightgbm · PyPI
- accelerate · PyPI
- tensorboard · PyPI
- groq · PyPI
- ray · PyPI
- torchvision · PyPI
- pytesseract · PyPI
- opt-einsum · PyPI
- weaviate-client · PyPI
- wandb · PyPI
- langchain · npm
- jax · PyPI
- tensorflow · PyPI
- browser-use · PyPI
- @huggingface/transformers · npm
- openai-agents · PyPI
- cohere · PyPI
- chromadb · PyPI
- llama-index · PyPI
- pymupdf-layout · PyPI
- ollama · npm
- sglang · PyPI