Anvil is an open-source, AI-native object store designed for modern workloads. We built it after hitting the limits of Git LFS, Hugging Face repos, S3, and others when working with multi-GB model files. It is S3-compatible & gRPC-native, supports: * Model-aware indexing - so it understands safetensors, gguf, and ONNX. * Tensor-level streaming * Erasure-coded storage * Open source (Apache-2.0) If you’re storing large models, versioning fine-tunes, running local inference, we want your feedback.
Hey all,
We’re the team behind Anvil.
We built this because every existing option failed when serving large models:
Git LFS broke on multi-GB safetensors
HF repos weren’t ideal for private/internal hosting
S3/MinIO treated model files as “dumb blobs”
Full-model downloads were too slow for inference
Replication made storage 3× the cost
Fine-tunes duplicated 10–20GB base models repeatedly
So we built Anvil as the object store we wish existed.
It’s S3-compatible, self-hosted, open-source, and understands ML model formats natively.
We run it in production on an 18-node cluster and are finally releasing it to the world.
Happy to answer every question — technical deep dives welcome!
If you like what we built, an upvote means the world to us ❤️
