AI models and papers trending on October 3, 2026
The edition of this day as it was recorded: models with the likes and downloads they had then, and the most upvoted papers of the daily list.
On Saturday, October 3, 2026 the model at the top of the ranking was Cloudflare/clef (image text to text), with 824 downloads in 30 days and 857 likes. 3 models were held back for having likes without downloads. The most upvoted paper (from the daily list of October 2, 2026) was "OneStreamer: Unifying Perception, Memory, and Proactive Response in Streaming Video Interaction" from Nanjing University, with 152 votes. 10 of the 15 leading papers had a code repository attached.
Recorded 2026-10-03 09:16 UTC · numbers are those of that moment
Papers
- 152
OneStreamer: Unifying Perception, Memory, and Proactive Response in Streaming Video Interaction
Nanjing UniversityarXiv 2610.01762 (opens in a new tab)Code (opens in a new tab)
- 130
On-Policy or Off-Policy Learning? A Systematic Study of Distillation Dynamics
University of CambridgearXiv 2609.35259 (opens in a new tab)
- 118
Adaptive Reward Routing: Dynamic Multi-Reward Optimization for Joint Audio-Video Diffusion via Forward-Process RL
- 71
Beyond Memory: Harnessing Long-Horizon Agents with Explicit Belief States
alibabaarXiv 2610.01415 (opens in a new tab)Code (opens in a new tab)
- 70
Hierarchical Continuous Diffusion Language Models
University of Illinois at Urbana-ChampaignarXiv 2610.02193 (opens in a new tab)Code (opens in a new tab)
- 69
Sharpening Tax in Post-Training
MetaarXiv 2610.01509 (opens in a new tab)Code (opens in a new tab)
- 68
Agent Priors-guided Policy Learning
National University of SingaporearXiv 2609.35690 (opens in a new tab)Code (opens in a new tab)
- 65
A Missing Piece for Trustworthy AI Reviewers: From Benchmarking Rhetorical Robustness to SciCore Review
- 65
World Observer: Joint Actor-Observer Generation for Persistent World Modeling
KAIST AIarXiv 2610.02162 (opens in a new tab)Code (opens in a new tab)
- 55
Transformers Stop Thinking Too Early, and a Tiny LoRA Fixes It
arXiv 2609.36585 (opens in a new tab)Code (opens in a new tab)
- 50
E-MoE: Enhanced Mixture-of-Experts for Non-Factorized Diffusion Language Models
- 50
ActiveSaddler: Automated Curriculum Learning for Agent Harness Optimization
- 50
ROWBench: Do Video Models Render What the Program Specifies?
Alaya LabarXiv 2610.02205 (opens in a new tab)Code (opens in a new tab)
- 43
Make Sparse Rewards Count: Density-Aware Reward Aggregation for Multi-Reward RL
kuaishou technologyarXiv 2610.00574 (opens in a new tab)Code (opens in a new tab)
- 43
AutoGUIWorld: Image Generators as Visual World Models for GUI Agent
Tencent HunyuanarXiv 2610.01215 (opens in a new tab)Code (opens in a new tab)
Models
- 1
Cloudflare/clef (opens in a new tab)
Image text to text · 824 downloads / 30 d · 857 likes
- 2
abenzerps/Qwen-Image-2.1-Uncensored-GGUF (opens in a new tab)
Text to image · 1.38M downloads / 30 d · 2,863 likes
- 3
Contrastive-LM/CLM-v0.1-8B (opens in a new tab)
Text ranking · 2,951 downloads / 30 d · 676 likes
- 4
Image to video · 1.63M downloads / 30 d · 6,026 likes
- 5
Qwen/Qwen-Image-2.1 (opens in a new tab)
Text to image · 81.7K downloads / 30 d · 2,853 likes
- 6
SupersonicLabs/Julia-1 (opens in a new tab)
Text classification · 2,909 downloads / 30 d · 376 likes
- 7
Image text to text · 6.9M downloads / 30 d · 16.8K likes
- 8
Cloudflare/clef-flash (opens in a new tab)
Image text to text · 1,303 downloads / 30 d · 291 likes
- 9
PSRben/VisionHOPE (opens in a new tab)
Image classification · 1,279 downloads / 30 d · 378 likes
- 10
Viggle/Qwen-Image-2.1-viggle-turbo (opens in a new tab)
Text to image · 241K downloads / 30 d · 545 likes
- 11
nvidia/Nemotron-3-Diarization (opens in a new tab)
Voice activity detection · 44.4K downloads / 30 d · 629 likes
- 12
orcarouter/OrcaSAQ-2-Cyber-27B-Uncensored-GGUF (opens in a new tab)
Text generation · 9,533 downloads / 30 d · 278 likes
- 13
prism-ml/Ternary-Bonsai-2-27B-gguf (opens in a new tab)
Text generation · 3.87M downloads / 30 d · 2,362 likes
- 14
TaichuAI/ZDTaichu5.0-9B (opens in a new tab)
Image text to text · 12.4K downloads / 30 d · 2,662 likes
- 15
ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-Coder-GGUF (opens in a new tab)
Image text to text · 261K downloads / 30 d · 208 likes
- convaiinnovations/laya · 5,019 likes, 0 downloads
- inclusionAI/Ming-Image-0.1-Design · 368 likes, 0 downloads
- lilylilith/QI_2.1_AnyAngle · 112 likes, 0 downloads
About the archive
How far back does the archive go?
Model lists exist from the first day the collector ran: Hugging Face does not publish past trending lists, so they cannot be rebuilt. Paper lists for earlier days were rebuilt from the Hugging Face daily papers archive, and those days are marked "papers only".
What is kept for each day?
The leading models with their task, likes and downloads on that day, the models held back, and the leading papers with their upvotes. Abstracts are not duplicated in the archive; each paper links to arXiv.
Do archived numbers change?
No. A day’s record is the last collection of that day and is not rewritten afterwards. Likes and downloads you see on an archived day are the values on that day, not today’s.
How are weeks defined?
ISO 8601 weeks: Monday to Sunday, numbered so that week 1 contains the first Thursday of the year. The address /week/2026-40 is week 40 of 2026.