Launch catalogue · 23 open-weight models

The models the EU operator will serve at mainnet · famous open-weight models that fit the phase-0 cluster

Install the CLI
23models in catalogue
EUjurisdiction at launch
Q3 2026public mainnet
Catalogue ingestion in progress. Manifests + chunks are being signed and pushed to eu.concordfaces.org. Status badges below update as each model lands.

Catalogue

# Model Pipeline Library License Status
Loading…

Install the concord CLI

Rust 1.80+ · macOS · Linux · Windows

install via cargo recommended
# Requires Rust 1.80+.
$ cargo install --git https://github.com/Concordfaces/concord concord-cli

$ concord --version
pre-built binaries linux · macOS · windows
# Linux / macOS — shasum-verified install
$ curl -fsSL https://concordfaces.org/install.sh | sh

# Windows (PowerShell)
PS> iwr https://concordfaces.org/install.ps1 -useb | iex

# From source
$ git clone https://github.com/Concordfaces/concord
$ cd concord && cargo build --release
$ ./target/release/concord --help

Pull a model

Verified ed25519 signature · content-addressed blake3 chunks

concord EU · DE
# Pull a model from the EU operator.
$ concord pull sentence-transformers/all-MiniLM-L6-v2

# Verify the manifest signature + every chunk digest.
$ concord verify sentence-transformers/all-MiniLM-L6-v2

# Pull a specific version (omit for latest signed).
$ concord pull BAAI/bge-small-en-v1.5:v1

# Show local cache.
$ concord list --cached
Default endpoint. The CLI talks to api.eu.concordfaces.org for the EU operator. Override with --endpoint https://api.<op>.concordfaces.org or set CONCORD_ENDPOINT. Cache lives at ~/.cache/concord (override with CONCORD_CACHE or $XDG_CACHE_HOME/concord).

Using the HuggingFace CLI & huggingface_hub

Drop-in compatibility · zero code changes for most pipelines

huggingface-cli via HF_ENDPOINT
$ pip install huggingface_hub
$ export HF_ENDPOINT=https://api.eu.concordfaces.org/hf
$ huggingface-cli download \
    sentence-transformers/all-MiniLM-L6-v2

# Files land in the standard HF cache layout.
#   ~/.cache/huggingface/hub/models--<org>--<name>/
python · huggingface_hub snapshot_download
# Either set the env var once …
$ export HF_ENDPOINT=https://api.eu.concordfaces.org/hf

# … or pass `endpoint=` explicitly per call.
>>> from huggingface_hub import snapshot_download
>>> snapshot_download(
...     "BAAI/bge-small-en-v1.5",
...     endpoint="https://api.eu.concordfaces.org/hf",
... )
Why use concord over huggingface-cli? The native client streams signed manifest envelopes and verifies every blake3 chunk against the operator's ed25519 key on the wire — the HF shim is byte-for-byte compatible but the signature check happens server-side. Use concord verify after an HF-style pull if you need end-to-end proof.
Residency stays observed either way. Cross-border pulls still require a signed cross-border token (concord token mint) and are logged per manifest by the serving operator, regardless of which client you use.