Search across 729 pages

Try a tool name, category, or "lifetime deal"

Live directory · Updated weekly · No pay-to-rank

Hugging Face Models: The Complete Directory & Live Ranking

A live directory of Hugging Face Hub models, ranked by real adoption, growth, and maintenance - not opinion.

300Models ranked
10107.9MCombined downloads
8Categories
Live data from the Hugging Face Hub API
Ranked by real adoption & growth
No model pays to be listed or ranked
See the full ranking methodology →
Updated July 15, 2026 Hugging Face Hub API 300 models ranked Refreshed weekly Independent & ad-free
What & how

What Is a Hugging Face Model, and How Do You Use One?

What it is

A Hugging Face model is a pretrained AI model (language, vision, speech, or multimodal) published on the Hugging Face Hub, along with a model card documenting its license, training data, and intended use. Anyone can download, fine-tune, or run these models, often for free.

How to use a Hugging Face model

Every model on this list can be loaded directly with the transformers library (AutoModel.from_pretrained("org/model-id")) or downloaded standalone with huggingface_hub's snapshot_download. Click a model's name to open its official Hugging Face page, where the model card documents exact usage, required libraries, and any license terms to accept.

Browse the full directory below ↓

As of July 15, 2026, this is a live, free directory ranking the best and most-adopted models on the Hugging Face Hub by real data. zplatform.ai tracks models that clear a minimum adoption floor (300 ranked today) using the free, public Hugging Face Hub API. Ranked by downloads, likes, and how recently each model was updated - not opinion - and the list refreshes automatically every week. How we choose and rank →

The best Hugging Face models as of July 15, 2026, ranked by our composite score:

  1. gemma-4-26B-A4B-it - 38.3M downloads, quality score 91/100.
  2. gemma-4-31B-it - 37.1M downloads, quality score 91/100.
  3. chronos-2 - 105.4M downloads, quality score 90/100.
  4. chronos-2 - 50.9M downloads, quality score 90/100.
  5. gemma-4-E4B-it - 20.3M downloads, quality score 90/100.

Every Ranked Model

Every Hugging Face model that clears our adoption floor, scored on a transparent composite of adoption, growth, maintenance, and trust. Select up to 4 to compare side by side.

Compare # Model Category Downloads Likes License Updated Quality
Editor's picks

Best Hugging Face Models by Use Case

Best Vision & Multimodal Models

gemma-4-26B-A4B-it and gemma-4-31B-it lead the vision-language models on Hugging Face, handling image understanding, visual question answering, and multimodal chat in one model.

  1. gemma-4-26B-A4B-it 91/100 quality
  2. gemma-4-31B-it 91/100 quality
  3. gemma-4-E4B-it 90/100 quality

Best Text Generation & Chat Models

phi-4 and DeepSeek-V4-Flash top the text-generation category, the models behind most chat assistants and text-completion tools built on Hugging Face.

  1. phi-4 87/100 quality
  2. DeepSeek-V4-Flash 86/100 quality
  3. DeepSeek-V4-Pro 86/100 quality

Best Tabular, Time-Series & Robotics Models

chronos-2 and chronos-2 lead this emerging mix of tabular, time-series forecasting, and robotics models.

  1. chronos-2 90/100 quality
  2. chronos-2 90/100 quality
  3. openvla-7b 80/100 quality

Best Speech & Audio Models

wav2vec2-large-xls-r-300m-Urdu and parakeet-tdt-0.6b-v2 top speech and audio, covering transcription, text-to-speech, and audio classification.

  1. wav2vec2-large-xls-r-300m-Urdu 87/100 quality
  2. parakeet-tdt-0.6b-v2 86/100 quality
  3. distil-large-v3 85/100 quality

Best Text Classification Models

roberta-base-go_emotions and deberta-v3-base-prompt-injection-v2 top text classification and understanding tasks like sentiment, masked-token prediction, and named-entity tagging.

  1. roberta-base-go_emotions 90/100 quality
  2. deberta-v3-base-prompt-injection-v2 85/100 quality
  3. multilingual-sentiment-analysis 83/100 quality

Best Embedding & Retrieval Models

Qwen3-Embedding-0.6B and all-MiniLM-L6-v2 lead embeddings and retrieval, turning text into vectors for semantic search, RAG pipelines, and re-ranking.

  1. Qwen3-Embedding-0.6B 90/100 quality
  2. all-MiniLM-L6-v2 90/100 quality
  3. MoLFormer-XL-both-10pct 86/100 quality

Best Image Generation Models

Juggernaut-XL-v9 and Z-Image-Lora lead text-to-image generation on Hugging Face.

  1. Juggernaut-XL-v9 83/100 quality
  2. Z-Image-Lora 81/100 quality
  3. moeFussion 79/100 quality
Browse

Browse by Category

Vision & Multimodal

136 models · 1745.0M downloads

Filter the directory to Vision & Multimodal ↓

Text Generation & Chat

62 models · 545.0M downloads

Filter the directory to Text Generation & Chat ↓

Embeddings & Retrieval

29 models · 6598.4M downloads

Filter the directory to Embeddings & Retrieval ↓

Speech & Audio

25 models · 418.4M downloads

Filter the directory to Speech & Audio ↓

Image Generation

4 models · 8.1M downloads

Filter the directory to Image Generation ↓

Text Classification & Understanding

6 models · 116.3M downloads

Filter the directory to Text Classification & Understanding ↓

Tabular, Time-Series & Robotics

6 models · 319.4M downloads

Filter the directory to Tabular, Time-Series & Robotics ↓

Other

32 models · 357.4M downloads

Filter the directory to Other ↓

State of the Models Ecosystem

How we rank, and why you can trust it

  • Data source: the free, public Hugging Face Hub API (downloads, likes, license, and last-modified date for every model). Refreshed weekly.
  • Inclusion is automatic: a model must clear an adoption floor - at least 1,000 downloads in the last 30 days, or at least 50 likes - so the list stays credible rather than padded with unused uploads. We do not hand-pick winners.
  • Adoption is measured by all-time downloads (when available) and likes, log-scaled so a handful of enormous outliers do not flatten the rest of the ranking.
  • Growth is measured by the 30-day change in downloads, compared against our own weekly snapshots - shown once enough history has accrued. Hugging Face also exposes an internal "trending score," but its scale is not publicly documented, so it is shown for reference only and does not drive the ranking.
  • Maintenance reflects how recently a model was updated - a model updated in the last 30 days scores highest, decaying the longer it has gone untouched.
  • Ranking is data-driven only. No model pays to be listed or ranked.

The Quality Score formula

The Quality Score (0-100) is a transparent weighted blend: 40% Adoption (downloads and likes, log-scaled), 20% Growth (30-day downloads trend), 25% Maintenance (how recently the model was updated), and 15% Trust (a stated license, an identified author, and not being access-gated).

Frequently asked questions

What are the best Hugging Face models?

The best Hugging Face models by our composite data score are listed in the leaderboard above, ranked on adoption, growth, and maintenance rather than opinion. Use the filters to narrow by task category.

How is this different from Hugging Face's own trending page?

Hugging Face's own trending view surfaces short-term spikes. This directory instead weighs sustained adoption (downloads and likes) alongside how actively a model is maintained, so the ranking favors models that are genuinely being used, not just currently trending.

Are Hugging Face models free to use?

Most models on the Hugging Face Hub are free and open-weight, though licenses vary (Apache 2.0, MIT, and custom licenses like Llama's are all common). Some models are "gated" and require accepting terms before download. Each row shows the license when Hugging Face reports one.

How often is this list updated?

This directory refreshes weekly, pulling fresh downloads, likes, and maintenance data directly from the Hugging Face Hub API.

How many models are on the Hugging Face Hub?

The Hugging Face Hub hosts well over a million models. This directory ranks the 300 that clear our adoption floor (at least 1,000 downloads in 30 days, or 50 likes), as of July 15, 2026.

Last updated: . Data from the Hugging Face Hub API. See the methodology and the machine-readable data.json.