safetensors

Powerful Powerful Machine Learning library with GPU, CPU and WASM backends

600 KB WASM runtime for Cactus Compute's Needle AI tool-calling models, Needle 3, 2 and 1 from one build. Browser, Cloudflare Workers, Node.js, Python, C FFI and no_std. Token-exact with the JAX reference. No backend, no API key.

A utility to inspect, validate, sign and verify machine learning model files.

Pure Rust DLPack implementation for zero-copy tensor interoperability

A fast Rust CLI for LLM model merging, diffing, and conversion. 10 merge algorithms, mergekit-compatible configs, safetensors native.

The all-purpose viewer for the AI era. A native workspace for code, documents, data, and Git. macOS + Windows.

Download, inspect, and compare HuggingFace models from Rust. Multi-connection parallel downloads plus safetensors header inspection via HTTP Range. No weight data downloaded.

A high-performance neural network library built with Rust and CUDA for blazing-fast deep learning

UMC — The ffmpeg of AI models. Convert any model format to any other format (GGUF, ONNX, SafeTensors, CoreML, TensorRT...) without quality loss, at maximum speed, with mathematical proof. Built in Rust.

A fast, ffprobe-like CLI for inspecting local and remote .safetensors files.

Rust-native AI inference & training engine. Block AttnRes linear-time transformer with profile-driven GPU dispatch, TurboQuant, GGUF/SafeTensors, LoRA, multimodal (vision/audio/video), FerrisRes Armor security. 39K LOC, 951 tests. Adapts to any GPU via wgpu — Intel iGPU to H100. No Python.

Native PP-OCRv6 inference in Rust with CPU and WGPU backends (community project)

A high-performance local LLM inference engine written in Rust. Dynamically streams transformer layers to break VRAM limitations and run massive models on consumer hardware.

Reverse-engineering analysis of hologram-ai: formal functions, Alloy models, exact arithmetic, executable oracle. AGPL-3.0-only.

Pure-Rust BERT-compatible text embedding inference for local-first applications.

Generate BERT-compatible text embeddings locally in Rust without Python or external ML runtimes using a simple, efficient library.

Fast, lightweight static text embeddings for Rust and Python

Efficient compression for neural network weights

Bakes a model's safetensors weights into standalone Rust const arrays for the morsel embedded-ML crate

⚡ Ultra-High-Throughput Storage Engine for AI & Tensors (100% Rust + Python). 16µs random lookups (854x faster than HDF5), 50% checkpoint deduplication (FastCDC), thread-per-core sharding & non-blocking SWMR. Drop-in replacement for SafeTensors.