This is the Hugging Face trending list as the API returned it on November 13, 2026 at 17:17 UTC. A community fine-tune of a 9B Qwen distillation was first, MiniMax's H3 image model was second, and seven of the top 20 repositories were built on a Qwen base model. Eleven of the top 20 were created in the previous two weeks.
Top 20 Trending Models
Downloads are the API's recent-window count, not all-time downloads. Size is the parameter total in the repository's safetensors or GGUF metadata, where the API exposes one.
| Rank | Repository | Task | Likes | Recent downloads | Created | License | Size |
|---|---|---|---|---|---|---|---|
| 1 | xtools-at/DeepSeek-V4-Pro-Qwen3.5-9B-UNCENSORED-heretic | Text generation | 4,896 | 21,586 | Oct 30, 2026 | Apache 2.0 | 9.41B |
| 2 | MiniMaxAI/MiniMax-H3-Image | Text to image | 971 | 437,039 | Nov 3, 2026 | MiniMax H3 Image community license | Not listed |
| 3 | NousResearch/Arete-27B | Text generation | 524 | 5,545 | Nov 6, 2026 | Apache 2.0 | 27.36B |
| 4 | Lightricks/LTX-2.5 | Image to video | 11,724 | 2,778,909 | Jul 23, 2026 | LTX-2.x community license, gated | Not listed |
| 5 | speridlabs/idol | Image to 3D | 2,265 | 31,367 | Oct 28, 2026 | Apache 2.0 | Not listed |
| 6 | Qwen/Qwen3.8-27B | Image and text to text | 19,825 | 6,960,395 | Aug 5, 2026 | Apache 2.0 | 27.78B |
| 7 | moonshotai/Kimi-K3.1-Pro | Image and text to text | 2,003 | 40,953 | Nov 2, 2026 | Kimi K3.1 Pro license | 2.8T total, 104B active (model card) |
| 8 | DavidAU/Qwen3.8-27B-TURBO-Fable-Cold-Fusion (GGUF, name shortened) | Image and text to text | 3,022 | 1,639,400 | Sep 1, 2026 | Apache 2.0 | 26.90B |
| 9 | Comfy-Org/MiniMax-H3-Image | Text to image | 330 | 1,257,301 | Nov 8, 2026 | MiniMax H3 Image community license | Not listed |
| 10 | ccore/Qwen3.5-4B-3D-Coder | Not set | 1,071 | 392,582 | Oct 25, 2026 | Apache 2.0 | 4.54B |
| 11 | black-forest-labs/flux-3-image-klein-comfy | Image to image | 3,070 | 1,330,289 | Oct 14, 2026 | FLUX community license, gated | Not listed |
| 12 | Zyphra/ZONOS3-Mini | Text to speech | 264 | 13,068 | Nov 9, 2026 | Apache 2.0 | 0.89B |
| 13 | VideoRebirthAI/hyperflow-h3-flash | Image and text to video | 251 | 389,065 | Nov 4, 2026 | MiniMax H3 community license | 13.95B |
| 14 | unsloth/Qwen3.8-27B-GGUF | Not set | 5,847 | 6,253,865 | Aug 13, 2026 | Apache 2.0 | 27.32B |
| 15 | Cactus-Compute/mirage | Image and text to text | 230 | 18,959 | Nov 5, 2026 | Apache 2.0 | Not listed |
| 16 | ricursiveAI/Ripple-65B-Preview | Not set | 547 | 8,500 | Nov 4, 2026 | Apache 2.0 | 64.78B |
| 17 | Viggle/Viggle-Remake | Video to video | 359 | 183,188 | Oct 22, 2026 | MiniMax H3 community license | 20.11B |
| 18 | mradermacher/gemma-4-31B-it-ARA-heretic-GGUF | Not set | 248 | 105,256 | Nov 7, 2026 | Tagged Apache 2.0, card links the Gemma 4 license | 30.70B |
| 19 | facebook/sam3 | Mask generation | 6,507 | 2,092,498 | Nov 7, 2025 | Custom, gated | 0.86B |
| 20 | tencent/Qwen3.8-27B-DSpine-b16 | Text generation | 223 | 4,332 | Nov 7, 2026 | Apache 2.0 | 2.03B in this repository, on a Qwen3.8-27B base |
What the List Shows
- Qwen is the base everyone builds on. Five of the top 20 are Qwen3.8-27B or built on it (ranks 3, 6, 8, 14 and 20), and two more are built on Qwen3.5 (ranks 1 and 10). In the top 30, nine repositories trace to Qwen3.8.
- Trending rewards new repositories. Eleven of the top 20 were created on or after October 30. The number one model had 21,586 recent downloads, while Qwen3.8-27B at rank six had 6.96 million.
- Half the list is not a chat model. Image, video, 3D, speech and segmentation models take nine of the top 20 places.
- Permissive licenses lead. Twelve of the top 20 carry an Apache 2.0 tag. The image and video models mostly ship under custom community licenses with revenue limits.
- Community edits outrank the originals. Four of the top 20 are modified or "uncensored" versions of a lab's model, including the model in first place.
What the Top Five Are
- DeepSeek-V4-Pro-Qwen3.5-9B-UNCENSORED-heretic: a community version, with refusals removed, of a 9B Qwen3.5 model that was distilled from DeepSeek V4 Pro on math and science samples.
- MiniMax-H3-Image: one checkpoint for text-to-image and editing with up to 16 reference images. The model card says it outputs up to 4096 by 4096 pixels in six sampling steps.
- Arete-27B: a Nous Research fine-tune of Qwen3.8-27B. The card reports its own 803-prompt evaluation, which no third party has reproduced.
- LTX-2.5: Lightricks' video and audio generation model with open weights, built to run locally.
- idol: turns a single image into a 3D Gaussian splat. The card says it needs 7.2 GB of VRAM.
Top Five Trending Datasets
| Rank | Dataset | Likes | Recent downloads | License |
|---|---|---|---|---|
| 1 | Angais/pixel-art-atlas-1.4m | 455 | 53,473 | CC0 1.0 |
| 2 | Vyber07/cyber-security | 1,005 | 20,450 | Apache 2.0, gated |
| 3 | nvidia/simready-assets | 1,566 | 1,432,334 | CC BY 4.0 |
| 4 | NousResearch/arete-data-pairs | 67 | 1,424 | No license tag |
| 5 | tencent/TailorRig | 93 | 5,714 | Custom |
Why Hugging Face Was in the News This Week
US searches for "hugging face" rose on November 12. That day Nvidia announced that Jensen Huang had stepped down as chief executive, according to CNBC. TechCrunch reported in September that Nvidia plans to acquire Hugging Face for $12.9 billion, with the deal expected to close in the first half of 2027. TechCrunch also reported that Hugging Face chief executive Clément Delangue responded publicly on November 12, saying the company would "keep fighting for open-source AI, harder than ever." The trending list above is not affected by that news.
Methodology
Rankings, likes, downloads, dates and license tags come from the Hugging Face models API sorted by trending score, read on November 13, 2026 at 17:17 UTC. The trending list changes hourly, so the live trending page will differ. We confirmed on one model that the download field is a recent-window count and not the all-time total. Model descriptions come from each model card and are the authors' own claims. Presenc AI did not test these models. For the earlier snapshot see June 2026.