Hugging Face Hub

Top Builders

Explore the top contributors showcasing the highest number of app submissions within our community.

Hugging Face Hub

The Hugging Face Hub is an open-source repository platform that hosts over one million machine learning models, datasets, and interactive applications. It serves as the central collaboration layer for the ML community, enabling developers to discover, share, version, and deploy models across every modality including text, vision, audio, and multimodal. Model checkpoints on the Hub are compatible with the Transformers, Diffusers, and Datasets libraries, and can be loaded in a few lines of code.

General
Author	Hugging Face
Type	ML Model Repository and Collaboration Platform
Website	huggingface.co
Documentation	Hub Documentation
Repository	github.com/huggingface/huggingface_hub
Models	huggingface.co/models
Datasets	huggingface.co/datasets

Start building with Hugging Face Hub

The Hub is the fastest way to get a pretrained model running in your project. Load any of the 1M+ checkpoints directly into Transformers or Diffusers with a single function call, or browse the Hub to find the right base model for your use case. You can host your own models privately and share fine-tuned adapters with the community without uploading full model weights. During AMD-sponsored hackathons on lablab.ai, participants pull models from the Hub, fine-tune or build on them using AMD Developer Cloud GPUs, and publish their final projects back to the Hub as a Space. Explore what the community has built at Hugging Face Use Cases and Applications.

Hugging Face Hub Tutorials

Getting Started

Hub Documentation Full reference for model cards, repos, and the Hub API
huggingface_hub Python library Python SDK for interacting with the Hub programmatically
Model Hub Browse all publicly available models by task, framework, and license
Datasets Hub Browse curated datasets for training and evaluation

Key Features

1M+ models Text, vision, audio, multimodal, and specialized domain models from top research teams and companies including Meta, Mistral, Google, and Alibaba Cloud.

Private repositories Host proprietary models and datasets with access controls. Upgrade to a PRO or Enterprise account for private inference endpoints.

Model cards Structured documentation for model limitations, intended use, training details, and evaluation results — standardized across all public checkpoints.

Version control Git-based versioning with LFS support for large files. Every model and dataset on the Hub has a full commit history.

Fine-tuned adapters Share and reuse LoRA and PEFT adapters without uploading full model weights. Adapters reference their base model and load in seconds.

Libraries

Transformers Unified API for pretrained models across text, vision, and audio
huggingface_hub Python SDK for Hub authentication, uploads, and downloads
Datasets Efficient access to Hub datasets with streaming and arrow-based caching
PEFT Parameter-efficient fine-tuning (LoRA, QLoRA, prefix tuning)
Optimum-AMD Optimized inference and training for AMD hardware via ROCm

Edit on GitHub

huggingface HuggingFace Hub AI technology Hackathon projects

Discover innovative solutions crafted with huggingface HuggingFace Hub AI technology, developed by our community members during our engaging hackathons.

Verdict — AI counterparty due diligence agent

Verdict is an autonomous due-diligence agent for business payments. Every B2B payment carries counterparty risk: shell companies, sanctioned entities, vendors with fraud histories. The evidence to catch them lives on the open web, but rate limits, bot detection, CAPTCHAs and geo-blocks defeat ordinary automation. Give Verdict a counterparty name, and it runs a live investigation — discover, access, extract — then returns a verdict: APPROVE, ESCALATE, or BLOCK, with a 0-100 risk score and a source citation behind every finding. It supports single checks and batch screening that ranks a whole vendor list, riskiest first. How it meets the criteria: Bright Data integration (load-bearing): Verdict cannot function without live web access. It uses the Bright Data MCP server — search_engine for risk-focused discovery, and scrape_as_markdown to access sites that block ordinary scrapers. Model integration: An LLM reads the sanitized evidence and returns a structured JSON verdict. Risk thresholds (0-25 APPROVE / 26-69 ESCALATE / 70-100 BLOCK) are enforced in code so score and verdict never disagree. Business impact: Counterparty screening is a control fintechs and finance teams pay for today. Verdict automates it for every transaction, not just big deals. Trust engineering: Scraped content is treated as untrusted and sanitized before reaching the model, mitigating prompt injection — essential for a system that gates money. Proven live: Apple Inc → APPROVE (low risk). Wirecard AG → BLOCK (risk 95/100), surfacing null-and-void financials, non-existent trust assets, and executives pursued by authorities — all cited to real sources. Deployed and usable online.

A-JEPA AUTOMATA

High Level Overview Automata is a production-grade AI agent platform that combines live web intelligence with formal verification — making it the first system where AI-driven business decisions are mathematically auditable before execution. The core problem: enterprise teams can't trust AI agents acting on web data because there's no proof the reasoning is sound. Automata solves this with a three-layer stack. Layer 1 — Web Intelligence Intake (Bright Data): The Bright Data MCP Server and Web Scraper API feed structured live data — competitor pricing, regulatory filings, LinkedIn hiring signals, SERP trends — directly into the ingestion pipeline. Web Unlocker handles bot-protected sources. All intakes logged to a Blake2b-hashed append-only audit trail from the first byte. Layer 2 — Agentic Orchestration: A FastAPI backend with async workers processes ingested signals. The Go CLI harness runs named analysis flows — sorry scan, interconnect map, signal diff — and exposes structured JSON for downstream AI agents. A proof watcher tracks theorem and proof-completion metrics per file in real time, ensuring that the logic layer never silently regresses. Layer 3 — Formal Verification : Every intelligence claim that triggers an action passes through an Automata state machine. The proof_completion metric — theorems minus sorry-count divided by theorem-count — gates whether a decision is certified or flagged for human review. No sorry-equivalent proof, no downstream action. This is provable trust, not probabilistic trust. Infrastructure: Docker Compose stack with Postgres, Redis, Alembic migrations, Grafana/Loki observability, nginx reverse proxy, and an inotify-based file watcher. Deployable on ROCm hardware. Track coverage: GTM Intelligence (competitor and buying-signal monitoring), Finance & Market Intelligence (pricing and filing pipelines), Security & Compliance (regulatory change detection with proof-gated alerts). A single coherent system spanning all three tracks.

OmniClaims Adjuster

OmniClaims Adjuster: El Futuro de la Liquidación de Siniestros En la actualidad, el procesamiento de reclamos de seguros es un proceso manual, lento y propenso a errores. OmniClaims Adjuster revoluciona el sector Insurtech mediante una arquitectura multi-agente totalmente autónoma construida sobre la familia de modelos Gemini 3.1 de Google. Diseñado para la AI Agent Olympics Hackathon, este sistema actúa como un ajustador de seguros experto. En lugar de depender de un solo modelo monolítico, el flujo de trabajo orquesta múltiples agentes especializados trabajando en paralelo y en tiempo real: 1. Agente de Extracción: Transforma las narrativas no estructuradas del cliente en datos estructurados estandarizados bajo esquemas estrictos de Pydantic. 2. Agente de Pólizas: Analiza los términos contractuales (PDFs) verificando límites, deducibles, exclusiones y coberturas con precisión milimétrica. 3. Agente de Visión (Daños): Aprovecha la multimodalidad nativa de Gemini 3.1 Pro para examinar fotografías de evidencias, evaluando la severidad y la congruencia del daño reportado. 4. Agente Antifraude: Detecta anomalías cruzando variables (ej. inconsistencias entre la historia del cliente y la evidencia visual) para emitir una puntuación de riesgo. 5. Agente Orquestador: Consolida todos los análisis en una decisión final holística (Aprobado, Rechazado o Revisión Manual). A nivel técnico, la plataforma cuenta con un backend en FastAPI y una interfaz Gradio con diseño premium glassmorphism. Priorizando la explicabilidad (AI Transparency), el sistema expone en la UI todo el Chain of Thought (Razonamiento) de los agentes. OmniClaims Adjuster no reemplaza al ajustador humano; lo empodera resolviendo automáticamente el 80% de los casos claros y entregando un dossier procesado de alta inteligencia para los reclamos complejos.

MediTriage Multi Agent symptom triage

MediTriage is an advanced, safety-first clinical decision support prototype built for the Collaborative Systems track. It demonstrates how multi-agent orchestration can outperform a single LLM prompt in high-stakes domains like healthcare triage—where correctness, safety, and accountability are critical. At its core, MediTriage simulates a team of specialized clinical reasoning agents, each responsible for a clearly defined task in the triage pipeline. These agents collaborate through a structured workflow powered by LangGraph, producing outputs that are not only more reliable but also explainable and modular.

Autonomous AI Agents for CRISPR Diagnostics

Our project addresses the critical gap in healthcare accessibility across Africa by leveraging advanced autonomous AI agents to automate and streamline CRISPR diagnostics and complex clinical workflows. Powered by state-of-the-art large language models like Meta-Llama-3-8B-Instruct, the system acts as an intelligent layer capable of parsing complex genetic data, validating clinical outputs, and providing real-time decision support for healthcare providers. By optimizing laboratory diagnostics and handling administrative clinical structures autonomously, the platform dramatically reduces latency times, eliminates manual errors, and guarantees high-precision medical analysis. This ensures that cutting-edge precision healthcare becomes fast, affordable, and scalable for underserved populations.