Hugging Face

Hugging Face (Twitter)

RT @Nik__V__: MapAnything V1.1 Release is live! 🚨

✅ Improved Checkpoints
✅ Model factory to test & train many models
✅ Profiling
✅ New COLMAP demos & voxelization tooling
✅ WAI format Benchmarking Data

Time to update the CVPR subs 😉

Comparisons to DA3 (1.1), Pi3X & more info in 🧵👇

‌Hugging Face (Twitter)

RT @RisingSayak: Two papers made it to #ICLR26. Happy 🤗

StructBench: arxiv.org/pdf/2510.05091
NoiseRefine: arxiv.org/pdf/2412.03895

Grateful to my collaborators and HF for letting me in on the fun 🚗

Hugging Face (Twitter)

RT @ben_burtenshaw: We got Claude to teach open models how to write CUDA kernels.

This blog post walks you through transferring hard capabilities (like kernel writing) between models with agents skills. Here's the process:

- get a powerful model (like Claude Opus 4.5 or OpenAI GPT-5.2) to solve a hard problem
- convert that trace into an agent skill
- transfer it to open-source, cheaper, or local model
- measure if it actually helps

We tested this on a gnarly task: writing CUDA kernels for diffusers. The results? Some open models saw +45% accuracy improvements with the right skill.

But the skill didn't help every model equally. Some even degraded performance, or used way more tokens. If you're transferring skills, you should evaluate.

We used upskill, a new tool for generating and evaluating agent skills. It works like this:

uvx upskill generate "write nvidia kernels" --from ./trace.md

Hugging Face (Twitter)

RT @halcyonrayes: 🤖 sam-3d-objects from @AIatMeta and trellis.2-4b from @Microsoft now exist on 3d-arena by @huggingface, one of the best benchmarks out there to compare object model generation quality!

both releases were two of the most impactful 3d-generation models to come in recent times!

go out and vote on your favourite 3d models at the link below.

delicious and exciting stuff coming up for 3d generation!

Hugging Face (Twitter)

RT @latkins: Today, we are releasing our first weights from Trinity-Large, our first frontier-scale model in the Trinity MoE family. American Made.

- Trinity-Large-Preview (instruct)
- Trinity-Large-Base (pretrain checkpoint)
- Trinity-Large-TrueBase (10T pre Instruct data/anneal)

Hugging Face (Twitter)

RT @tbpn: BREAKING: Hugging Face CEO @ClementDelangue and his wife have welcomed twin daughters 🤗🤗

Hugging Face (Twitter)

RT @allen_ai: Introducing Ai2 Open Coding Agents—starting with SERA, our first-ever coding models. Fast, accessible agents (8B–32B) that adapt to any repo, including private codebases. Train a powerful specialized agent for as little as ~$400, & it works with Claude Code out of the box. 🧵

Hugging Face (Twitter)

RT @Xianbao_QIAN: Super excited to see the new Kimi K2.5 model released, with the usual congratulations from @Hailuo_AI

- Continuous pretrained on 15T visual & text token on top of Kimi K2
- Multimodal model with both image and *video* understanding
- Interleaved Thinking and Multi-Step Tool Call
- Both instant mode and thinking mode available with one config flip
- Parallel sub-task support https://twitter.com/Kimi_Moonshot/status/2016024049869324599#m

Hugging Face (Twitter)

RT @RisingSayak: Meet, greet, and grill @ariG23498 and me at the @PyTorch Day India, happening on 7th Feb, 2026, Bangalore.

It will be a technical talk -- just in case the title didn't suggest that already.

See you there.

Hugging Face (Twitter)

RT @allen_ai: Molmo 2 (8B) is now available via @huggingface Inference Providers, courtesy of Public AI.

State-of-the-art video understanding with pointing, counting, & multi-frame reasoning. Track objects through scenes and identify where + when events occur. 🧵

‌Hugging Face (Twitter)

RT @vanstriendaniel: Want $20 to try GLM-4.7-Flash?

Join toad-hf-inference-explorers on @huggingface - $20 free to run open models via Inference Providers.
GLM, DeepSeek, Qwen and 100s more.

Only 85 people have claimed it so far 🤷
http://huggingface.co/toad-hf-inference-explorers

Hugging Face (Twitter)

RT @LysandreJik: Transformers v5's FINAL, stable release is out 🔥 Transformers' biggest release.

The big Ws of this release:
- Performance, especially for MoE (6x-11x speedups)
- No more slow/fast tokenizers -> way simpler API, explicit backends, better performance
- dynamic weight loading: way faster, and enabling: MoE now working w/ {quants, tp, peft, ...}

We have a migration guide on the main branch; please take a look at it in case you run into issues. Come in our GH issues if you still do after reading it 😀

Hugging Face (Twitter)

RT @nvidianewsroom: 🌍 Weather forecasting has always relied on powerful supercomputers running physics-based models.

We are proud to announce the NVIDIA Earth-2 family of open models — the world’s first fully open, accelerated AI weather stack — saving computational time and costs, and enabling more nations, enterprises, and businesses to run application-specific forecasting systems.

Weather AI is now accessible worldwide at every stage. #AMS2026

Read more: nvda.ws/4sWQ2B4

Hugging Face (Twitter)

RT @overworld_ai: Step in, move around, and see the world update as you act.

We’ve put up a live Hugging Face demo of our real-time world model so you can check it out.

Hugging Face (Twitter)

RT @ltx_model: 2,000,000 @huggingface downloads!

LTX-2 reached this milestone the way we believe it should. Built in the open, shaped by real-world use, and driven by the community.

Thank you to everyone experimenting, building, and pushing it forward.

Looking ahead to what’s next.

‌Hugging Face (Twitter)

RT @RisingSayak: Some notes on what someone can do for building chops in ML x open source x modeling:

• Take a popular pre-trained model implementation, profile it, spot the bottlenecks, & try to improve its speed-memory trade-off -- it's a valuable skill that any sane hiring manager should understand and credit (they are probably not legit if they don't). is a good example of this.

• GPUs are in short supply. So, try reimplementing it in JAX, leveraging its strengths. Make it run on TPUs, blazing fast 🔥 -- this will help you establish that you care about performance and are comfortable switching stacks when needed. https://github.com/sanchit-gandhi/whisper-jax is an amazing example of this.

• In the context of an organization, communication is the key. Make sure you document your experience in an easily digestible way so most folks would understand what you achieved. Provide numbers on benchmarks, mention assumptions, and whatever limitations you faced and how you approached them.

• Get a pro subscription to whatever AI coding assistant you think works the best for your stuff. Make it a part of your workflow, but DO NOT become overly reliant on it. Have enough juice in the process so that you can build muscle memory and objective evidence of your intellect over time.

• Have fun!

It gives me a sense of joy and relief to know that back in the days, we did all of it happily WITHOUT any AI coding assistance. Lots of fun, despair, and anxiety -- but all worth it; 10/10 -- would do it again!

Hugging Face (Twitter)

RT @ariG23498: We (/w @RisingSayak) are going to represent @huggingface at the PyTorch Day India event this year.

Here is the plan, we go to Bengaluru, talk about Transformers and Kernels, coversate with like minded folks and then leave. Short and sweet!

If you are still on the edge of deciding whether to attend or not, we give you another good reason to attend. 🤗

See you there.

‌Hugging Face (Twitter)

RT @overworld_ai: If you just want to try it:

You can download the model here right now:
https://huggingface.co/Overworld/Waypoint-1-Small