Hugging Face's logo

Hugging Face Smol Models Research
Enterprise
community
Activity Feed

AI & ML interests

Exploring smol models (for text, vision and video) and high quality web and synthetic datasets

Recent Activity

thomwolf  authored a paper about 1 month ago
FineWeb2: One Pipeline to Scale Them All -- Adapting Pre-Training Data Processing to Every Language
lvwerra  authored a paper about 1 month ago
FineWeb2: One Pipeline to Scale Them All -- Adapting Pre-Training Data Processing to Every Language
craffel  authored a paper about 1 month ago
FineWeb2: One Pipeline to Scale Them All -- Adapting Pre-Training Data Processing to Every Language
View all activity

loubnabnl 
in HuggingFaceTB/SmolLM3-3B 2 days ago

Embed and add reference to EU transparency template

1
#29 opened 2 days ago by
yjernite
BrigitteTousi 
posted an update 2 days ago
view post
Post
259
This is what Hugging Face is all about. We want everyone, hobbyists, researchers and industry alike, to be able to contribute to AI because everyone is affected by it. Kudos to HF's @irenesolaiman for spreading the word!🔥🤗
Xenova 
posted an update 3 days ago
view post
Post
1699
Introducing Voxtral WebGPU: State-of-the-art audio transcription directly in your browser! 🤯
🗣️ Transcribe videos, meeting notes, songs and more
🔐 Runs on-device, meaning no data is sent to a server
🌎 Multilingual (8 languages)
🤗 Completely free (forever) & open source

That's right, we're running Mistral's new Voxtral-Mini-3B model 100% locally in-browser on WebGPU, powered by Transformers.js and ONNX Runtime Web! 🔥

Try it out yourself! 👇
webml-community/Voxtral-WebGPU
andito 
posted an update 4 days ago
view post
Post
2571
Many VLMs claim to process hours of video. But can they follow the story?🤔
Today, we introduce TimeScope: The benchmark that separates true temporal understanding from marketing hype. Let's see how much VLMs really understand!⏳

We test three skills that matter for real-world use:
🔎 Localized Retrieval: Find a specific action.
🧩 Information Synthesis: Piece together scattered clues.
🏃 Fine-Grained Perception: Analyze detailed motion (e.g., count how many times a person swings an axe).

The results are in, and they're revealing. Only Gemini 2.5 pro handles 1-hour-long videos.
Performance drops sharply with duration, proving that long video understanding is still challenging. We've found the breaking points—now the community can start fixing them.📈

Want to learn more? TimeScope is 100% open-source. Benchmark your model and help us build the next generation of video AI.

📖 Blog:
https://huggingface.co/blog/timescope-video-lmm-benchmark
👩‍💻 Leaderboard & Demo: Apollo-LMMs/TimeScope
📊 Dataset: Apollo-LMMs/TimeScope
⚙️ Eval Code: https://github.com/EvolvingLMMs-Lab/lmms-eval
merve 
posted an update 4 days ago
view post
Post
617
so many open LLMs and image LoRAs dropped past week, here's some picks for you 🫡 merve/releases-july-18-687e3fbd2ab9b39c51f9238b

LLMs
> ByteDance released a bunch of translation models called Seed-X-RM (7B) ByteDance-Seed/Seed-X-RM-7B
> NVIDIA released reasoning models of which 32B surpassing the giant Qwen3-235B with cc-by-4.0 license 👏 nvidia/openreasoning-nemotron-687730dae0170059860f1f01
> LG released a new EXAONE model (32B) LGAI-EXAONE/EXAONE-4.0-32B

VLMs/any-to-any
> vidore/colqwen-omni-v0.1 is a new any-to-any retriever (MIT)
> HiDream-ai/HiDream-E1-1 is image+text in image+text out model (MIT)

LoRAs
> There's a bunch of LoRAs based on Flux Kontext, gotta check out the collection 🤠
eliebak 
posted an update 6 days ago
view post
Post
4366
Kimi K2 tech report is full of gems as always. Here are my notes on it:

> MuonClip: Pretty crazy how after 70k the training stabilizes and the QK-clip is basically inactive. There is also no loss in perf with QK-clip which is not trivial at all (at small scale but with aggressive threshold). Also a cool explanation of why muon makes the logit explode in appendix E (tl;dr is that muon makes the singular value of the update matrix higher)
> Sparsity scaling laws to justify their ratio, they have a very solid training infra that allows the model to be trained at this sparsity level, they could have increased even more but as sparsity increases the training becomes less efficient.
> They diminish the number of attention heads to make it more efficient for long context since attention heads are a big bottleneck for long context. They also remove 2 of the 3 "first dense" layers in the dsv3 arch.

With the sparsity and attention heads (divided by 2) they achieve 83% increased flops compared to deepseek v3 arch at 128k.

> Data: Rephrasing is KEY. They do a lot more synthetic data generation and rephrase their corpus to have different styles, for longer documents they do it by chunk. I'm (half) surprised by the fact that ONLY 1 epoch (assuming same number of training tokens I think?) of data rephrased 10 times has better accuracy than 10 epochs of the same data rephrased once.
> They do rewriting for Math and Knowledge, for Math they apply the ShallowMath recipe and instruct the model to rephrase in a "learning note" style
> They talk about diversity and probably have some internal stuff/eval to test that, as always still a bit unclear for me how to properly measure that.

The infra is also very nice, quick summary:
> PP=16 (1F1B schedule, a bit custom), EP=16, zero1
> No FP8 computation but for storage of specific layers, selective recomputation for inexpensive block, activation offloading to CPU
merve 
posted an update 6 days ago
fdaudens 
posted an update 9 days ago
view post
Post
2026
AudioRAG is becoming real! Just built a demo with ColQwen-Omni that does semantic search on raw audio, no transcription needed.

Drop in a podcast, ask your question, and it finds the exact chunks where it happens. You can also get a written answer.

What’s exciting: it skips transcription, making it faster and better at capturing emotion, ambient sound, and tone, surfacing results text search would miss.

- Demo: fdaudens/colqwen-omni-demo
- Blog post from ColQwen team: https://huggingface.co/blog/manu/colqwen-omni-omnimodal-retrieval
  • 1 reply
·
merve 
posted an update 10 days ago
merve 
posted an update 11 days ago
view post
Post
2545
Fine-tune Gemma3n on videos with audios inside with Colab A100 🔥
Just dropped the notebook where you can learn how to fine-tune Gemma3n on images+audio+text at the same time!

keep in mind, it's made for educational purposes 🫡 we do LoRA, audio resampling & video downsampling to be able to train <40GB VRAM

stretch modalities and unfreeze layers as you wish! 🙏🏻 merve/smol-vision
  • 1 reply
·
fdaudens 
posted an update 12 days ago
view post
Post
2460
You might not have heard of Moonshot AI — but within 24 hours, their new model Kimi K2 shot to the top of Hugging Face’s trending leaderboard.

So… who are they, and why does it matter?

Had a lot of fun co-writing this blog post with @xianbao , with key insights translated from Chinese, to unpack how this startup built a model that outperforms GPT-4.1, Claude Opus, and DeepSeek V3 on several major benchmarks.

🧵 A few standout facts:

1. From zero to $3.3B in 18 months:
Founded in March 2023, Moonshot is now backed by Alibaba, Tencent, Meituan, and HongShan.

2. A CEO who thinks from the end:
Yang Zhilin (31) previously worked at Meta AI, Google Brain, and Carnegie Mellon. His vision? Nothing less than AGI — still a rare ambition among Chinese AI labs.

3. A trillion-parameter model that’s surprisingly efficient:
Kimi K2 uses a mixture-of-experts architecture (32B active params per inference) and dominates on coding/math benchmarks.

4. The secret weapon: Muon optimizer:
A new training method that doubles efficiency, cuts memory in half, and ran 15.5T tokens with zero failures. Big implications.

Most importantly, their move from closed to open source signals a broader shift in China’s AI scene — following Baidu’s pivot. But as Yang puts it: “Users are the only real leaderboard.”

👇 Check out the full post to explore what Kimi K2 can do, how to try it, and why it matters for the future of open-source LLMs:
https://huggingface.co/blog/fdaudens/moonshot-ai-kimi-k2-explained
fdaudens 
posted an update 13 days ago
view post
Post
255
AI is reshaping everything—how we work, how we feel, even how nations compete.

Today’s reads cut across power, emotion, and disruption.

Here’s what stood out and why it matters 👇

AI might “solve” loneliness, but this could be a problem, as the discomfort of loneliness shapes us in important ways. 💔 https://t.co/k2Q9le6G0P

A new study warns of significant risks in using AI therapy chatbots, highlighting issues like stigmatization and inappropriate responses. 🤖 https://t.co/EFyW0RbYVl

AI is already showing signs of slashing job openings in the UK, particularly in roles exposed to the technology, suggesting a labor market slowdown. 📉 https://t.co/hhs0BbqIMa

AI firms like OpenAI are poaching Wall Street quants with massive paydays, shifting the talent landscape for building artificial general intelligence. 💰 https://www.businessinsider.com/ai-talent-openai-wall-street-quant-trading-firms-2025-7

Speaking of which: Nvidia CEO Jensen Huang disagrees with Anthropic CEO Dario Amodei on whether AI will create more jobs—or trigger a “white-collar apocalypse.” Huang believes AI will create vastly more, and better, jobs. ⚔️ https://t.co/YHWhY7qvSq

Can Nvidia convince governments to pay for “sovereign AI”? Politicians are warming to the idea of national AI systems, but it might not reduce dependence on US tech. 🌍 https://t.co/htQDzJAIDu
merve 
posted an update 13 days ago
view post
Post
2393
past week had huuuge releases 💗
here's our picks 🔥 find more models, datasets, demos here merve/releases-july-11-68750452c358c98b0fa663f7

> moonshotai/Kimi-K2-Instruct is the new sota LLM with 1T total 32B active parameters 🤯

> HuggingFaceTB/SmolLM3-3B is the new best LM for it's size, offers thinking mode 💭 as well as the dataset HuggingFaceTB/smoltalk2

> Alibaba-NLP/WebSailor-3B is the new agentic LLM for complex browsing

> Google DeepMind released medical vision LMs with an agentic doctor-patient app google/medgemma-release-680aade845f90bec6a3f60c4

> fal released a LoRA to improve details on face images fal/Realism-Detailer-Kontext-Dev-LoRA
merve 
posted an update 18 days ago
view post
Post
3088
GitHub refuses to render notebooks for a long time now 💔

so smol-vision now lives in Hugging Face model repository 🤗 merve/smol-vision
  • 1 reply
·
merve 
posted an update 19 days ago
view post
Post
3422
ByteDance released Tar 1.5B and 7B: image-text in image-text out models, fully open-source 👏 ByteDance-Seed/tar-6864cf0d9fe59a3b91cc4260

They have an image tokenizer unified with text, and they de-tokenize using either of two models (LLM and diffusion)
The model is actually a full LLM (Qwen2), the tokenizer converts image tokens 🤯
merve 
posted an update 20 days ago
view post
Post
3673
Huge drops in open AI past week!
Find more models, datasets, demos here merve/releases-july-4-686bcc54ed7c45c341fbf654
Some of our picks 🫡
⏯️ BAAI/MTVCraft is a new Veo3-like text-to-video model, demo is here BAAI/MTVCraft
🧑🏻‍💻 apple/diffucoder-6868139f56672ae046fe04e8 is a new family of diffusion LLMs (7B base and instruct) for coding
🗣️ kyutai/tts-1.6b-en_fr is a new small TTS model for English and France
👀 aharley/alltracker is a new pixel tracking model by Stanford, demo is here aharley/alltracker
📖 racineai/OGC_MEGA_MultiDomain_DocRetrieval is a new large visual document retrieval dataset
  • 1 reply
·
andito 
posted an update 24 days ago
view post
Post
3952
🧠👁️ Can AI visualize solutions?

Humans often solve visual problems by sketching ideas in our minds. What if Vision-Language Models (VLMs) could do something similar, not by generating full images, but by using internal “mental sketches”?

That’s the idea behind Mirage, a new framework that empowers VLMs to reason using latent visual tokens. Instead of just thinking in words, Mirage mixes in abstract visual representations that help the model solve complex tasks.

These aren't photorealistic images. They're compact, internal representations optimized purely to support reasoning.

🔧 Mirage is trained in two phases:

1) Grounding: It learns to produce latent tokens anchored in real images.
2) Refinement: The model drops the images and learns to generate visual tokens on its own.

📈 And yes, it works!
On challenging benchmarks like Visual Spatial Planning, Jigsaw puzzles, and Spatial Attention Tasks, Mirage clearly outperforms GPT-4o and other strong baselines.
Smart sketches > empty words.

By mimicking the way humans visualize solutions, Mirage gives AI a new kind of imagination, one that’s faster, more efficient, and more human-like.
Kudos to the teams at UMass Amherst and MIT behind this exciting work.
Check the paper: Machine Mental Imagery: Empower Multimodal Reasoning with Latent Visual Tokens (2506.17218)
·
merve 
posted an update 25 days ago
view post
Post
947
SOOOO MANY MODEL RELEASES 😍
Here's some picks from past week 🤗

> ByteDance/XVerse is a new identity preserving image generation model 🖼️
> google/gemma-3n-E4B-it, any-to-text model supported by transformers 🤗
> nvidia/llama-nemoretriever-colembed-3b-v1 two new state-of-the-art visual document retrievers 📑
> New version of Dia TTS model is up nari-labs/Dia-1.6B-0626
> Black Forest Labs releases Kontext benchmark black-forest-labs/kontext-bench

Find more here merve/releases-june-27-6864e8eb17f7e3a8b444083c
merve 
posted an update 25 days ago
merve 
posted an update 27 days ago