· Daily edition
Free to read · No sign-up · One edition per day
Daily edition · week 43 of 2026
Collected · snapshot shipped with the site build, not a live feed
A free daily front page of open AI: the models people are downloading most on Hugging Face and the papers researchers vote up, each with the plain facts you need to judge it.
Model of the day · No. 1Image text to text
deepseek-ai/DeepSeek-V4.1-Flash-1017 is an image text to text model published by deepseek-ai on Hugging Face. At the collection of October 24, 2026 it was No. 1 on the trending list, with 57,715 downloads in 30 days and 2,617 likes.
repository created 7 days ago; +18 likes since 2026-10-23.
MIT· commercial use allowed
Tianming Liang, Junyi Pan, Yuxi Li and 5 more · Tencent
arXiv 2610.31329 (opens in a new tab)PDF (opens in a new tab)Code (78 stars) (opens in a new tab)Repository on GitHub Trending Today (opens in a new tab)
Songlin Yang, Zhe Wang, Xiaotong Zhao and 8 more · Tencent
arXiv 2610.30926 (opens in a new tab)PDF (opens in a new tab)Code (1 stars) (opens in a new tab)Repository on GitHub Trending Today (opens in a new tab)
Xiaomin Yu, Hongliang Chen, Zhe Huang and 22 more
arXiv 2610.30181 (opens in a new tab)PDF (opens in a new tab)Code (23 stars) (opens in a new tab)Repository on GitHub Trending Today (opens in a new tab)
Wen Xue, Yong Liu, Xueheng Li and 5 more · jingdong
arXiv 2610.24738 (opens in a new tab)PDF (opens in a new tab)
A. Jian Yang, An Huang, Bo Dan and 174 more · IQuest
arXiv 2610.30949 (opens in a new tab)PDF (opens in a new tab)
Text generation39.8B parameters
No. 2 by Hugging Face trending score (683) · repository created 2 days ago · +15 likes since 2026-10-23.
1,904 downloads in 30 days · 704 likes
Apache-2.0· commercial use allowedImage to image
No. 3 by Hugging Face trending score (624) · repository created 9 days ago · +8 likes since 2026-10-23.
535K downloads in 30 days · 1,430 likes
Other (custom terms)· read the termsText to image5.08B parameters
No. 4 by Hugging Face trending score (444) · repository created 13 days ago · +4 likes since 2026-10-23.
7,658 downloads in 30 days · 660 likes
Apache-2.0· commercial use allowedImage to video
No. 5 by Hugging Face trending score (408) · +13 likes since 2026-10-23 · 2,026,951 downloads in the last 30 days.
2.03M downloads in 30 days · 9,066 likes
Other (custom terms)· read the termsText generation4.11B parameters
No. 6 by Hugging Face trending score (388) · repository created 5 days ago · +3 likes since 2026-10-23.
1,060 downloads in 30 days · 422 likes
Apache-2.0· commercial use allowedText to speech35.9B parameters
No. 7 by Hugging Face trending score (364) · +2 likes since 2026-10-23 · 17,017 downloads in the last 30 days.
17K downloads in 30 days · 955 likes
Other (custom terms)· read the termsText to image5.08B parameters
No. 8 by Hugging Face trending score (330) · repository created 13 days ago · +2 likes since 2026-10-23.
5,618 downloads in 30 days · 524 likes
Apache-2.0· commercial use allowedText to image
No. 9 by Hugging Face trending score (317) · +5 likes since 2026-10-23 · 2,398,071 downloads in the last 30 days.
2.4M downloads in 30 days · 4,917 likes
Other (custom terms)· read the termsImage text to text27.8B parameters
No. 10 by Hugging Face trending score (314) · +7 likes since 2026-10-23 · 7,710,011 downloads in the last 30 days.
7.71M downloads in 30 days · 18.3K likes
Apache-2.0· commercial use allowedNo task tag
No. 11 by Hugging Face trending score (306) · repository created 7 days ago · +4 likes since 2026-10-23.
1.61M downloads in 30 days · 405 likes
Apache-2.0· commercial use allowedThe order above is Hugging Face’s trending order at the time of collection. We do not score models; we only take out the entries below and say why.
Source: Hugging Face public API · updated
A model with 100+ likes and fewer than one download per 10 likes, 3 or more days after creation, leaves the ranking and is listed apart with its numbers. In this edition that is 0 of 40.
Each day is archived at /YYYY-MM-DD and each ISO week at /week/YYYY-WW. Hugging Face shows only the current list.
No machine summary is generated or shown. Past the leading papers the abstract is cut, marked as an excerpt, and linked to arXiv.
Each model shows its licence tag and what that licence family means for commercial use, next to parameter count and file formats.
You see when the numbers were read and whether they come from the scheduled collector or from the snapshot shipped with the build.
When Hugging Face lists a repository for a paper, the link is shown with its star count and a link to the same repository on GitHub Trending Today.
Each edition is a snapshot of two public lists taken at the time printed under the headline: the models at the top of Hugging Face’s trending list, and the most upvoted papers of the latest daily list. Nothing here is written by a language model, and nothing is reordered by taste.
Under every model name you will find the facts behind its position: its place and score on Hugging Face, how old the repository is, and, from the second day a model is listed, how many downloads and likes it gained since the previous edition. Hugging Face does not publish how the trending score is computed, so the line reports what can be observed and stops there.
A like is one click by a signed-in user. A download means someone fetched the weights. Read together they separate a model people are running from one people are only talking about. When the two disagree by an order of magnitude, the model leaves the ranking and appears in the box above with both numbers, so you can judge the case yourself.
Hugging Face compiles a paper list on working days. This edition carries the list of October 23, 2026. Votes keep arriving for a day or two after a list is published, so a paper’s count in the archive is the count on the day it was recorded.
From two public Hugging Face API addresses: the model list sorted by trending score, and the daily papers list. A scheduled job on our server reads them and stores the result; your browser only talks to this site. The time of the collection is printed under the headline of every page.
It is Hugging Face’s order. We do not compute a score. The single change we make is to move some models out of the ranking into a "held back" list. A model is held back when it has at least 100 likes, was created 3 or more days before the collection, and its downloads in the last 30 days are fewer than one per 10 likes. Each held-back entry shows the numbers that triggered the rule and its original position.
Likes cost nothing and can be organised; downloads require someone to actually fetch the weights. Hugging Face’s trending score reacts to recent likes, so a repository can reach the top of the list without being used. We cannot tell why it happened in a given case, and a model may be distributed outside Hugging Face, so we only report the two numbers side by side.
The collector is scheduled every three hours, starting at 00:07 UTC, so the edition of the day is refreshed through the day. If a collection fails, the previous edition stays on the page with its own date, and a notice appears once it is more than 36 hours old.
Hugging Face publishes a daily papers list on working days only. On weekends and holidays the edition carries the most recent list, looking back up to 4 days, and the papers section states which day that list belongs to.
No. The text under each paper is the abstract written by its authors, as published on arXiv and relayed by Hugging Face, and no machine summary replaces it. A wrong summary of a scientific paper is worse than none, so we do not write one. The only generated text about what a paper says is a short note on its technical words, labelled as automatic and shown directly under the abstract.