Where to Find the Official MTPLX Model Catalog on Hugging Face

The official MTPLX model catalog is hosted on Hugging Face under the MTPLX organization at https://huggingface.co/mtplx and defined in the repository's mtplx/model_catalog.py file.

The youssofal/MTPLX repository maintains a comprehensive model catalog that maps supported identifiers to Hugging Face repository URLs. This catalog serves as the single source of truth for MTPLX-supported models, with the data mirrored on Hugging Face for easy discovery and programmatic access.

Official Hugging Face Locations

The MTPLX project publishes its model catalog in two primary locations on the Hugging Face Hub. The main organization page provides a browsable index of all available models, while the dedicated catalog repository hosts structured metadata files.

• Organization Page: Visit https://huggingface.co/mtplx to browse all MTPLX models and associated repositories.
• Model Catalog Repository: Access the structured catalog data at https://huggingface.co/mtplx/model-catalog, which contains the official JSON manifest mapping model IDs to specific Hugging Face repo URLs.

Source Code Implementation

According to the youssofal/MTPLX source code, the catalog originates from local Python definitions before being synchronized to Hugging Face.

Catalog Definition

In mtplx/model_catalog.py, the catalog defines RAM-aware recommendation tiers that classify models by their memory requirements. This file maintains the authoritative mapping of model identifiers (such as mtplx/qwen3.5-27b-mtplx) to their corresponding Hugging Face repository paths. The data structures here serve as the ground truth for all model lookups within the MTPLX ecosystem.

Manifest Generation

The scripts/gen_models_manifest.py utility reads the Python catalog definitions and generates a JSON manifest suitable for publishing. This script transforms the internal model registry into a machine-readable format that the Hugging Face Hub consumes, ensuring the online catalog stays synchronized with the codebase.

Programmatic Access

You can interact with the MTPLX model catalog using the huggingface_hub library or the MTPLX command-line interface.

Listing Models via Hugging Face Hub

Use the huggingface_hub Python library to enumerate all MTPLX organization models:

from huggingface_hub import HfApi

api = HfApi()
mtplx_models = api.list_models(author="mtplx")

for model in mtplx_models:
    print(f"{model.modelId}: {model.tags}")

Pulling Models with MTPLX CLI

For direct model downloads, use the MTPLX command-line tool which references the catalog definitions:

import subprocess

model_id = "mtplx/qwen3.5-27b-mtplx"
subprocess.run(["mtplx", "pull", model_id], check=True)

Dependencies and Version Requirements

The MTPLX codebase specifies strict dependency requirements for Hugging Face integration.

In pyproject.toml, the project declares huggingface-hub>=0.36 as a core dependency for catalog access. The uv.lock file further pins specific versions of the Hugging Face Hub library to ensure consistent API compatibility when fetching model metadata from the official catalog endpoints.

Summary

• The official MTPLX model catalog on Hugging Face is accessible at https://huggingface.co/mtplx and https://huggingface.co/mtplx/model-catalog
• Source definitions reside in mtplx/model_catalog.py with RAM-aware configuration tiers
• The scripts/gen_models_manifest.py script generates the published manifest for downstream tooling
• Requires huggingface-hub>=0.36 as specified in pyproject.toml and locked in uv.lock

Frequently Asked Questions

Where is the official MTPLX model catalog hosted on Hugging Face?

The catalog is hosted under the MTPLX organization at https://huggingface.co/mtplx, with structured metadata available in the model-catalog repository at https://huggingface.co/mtplx/model-catalog. These locations contain all supported model IDs and their corresponding repository mappings according to the source code registry.

How does the MTPLX source code define the model catalog?

The catalog is defined in mtplx/model_catalog.py, which specifies RAM-aware recommendation tiers and maps model identifiers to Hugging Face URLs. This file serves as the authoritative source that gets synchronized to the Hugging Face organization via the manifest generation script.

What script generates the MTPLX model manifest for Hugging Face?

The scripts/gen_models_manifest.py utility processes the local catalog definitions in model_catalog.py and generates a JSON manifest. This manifest is then published to the Hugging Face catalog repository, ensuring the online listing matches the source code for downstream consumption.

What version of huggingface-hub is required to access the MTPLX catalog?

The pyproject.toml file specifies huggingface-hub>=0.36 as the minimum required version. The uv.lock file locks specific compatible versions to ensure reliable programmatic access to the catalog endpoints without API conflicts.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}