LLMs No Longer Require Powerful Servers: Researchers from MIT, KAUST, ISTA, and Yandex Introduce a New AI Approach to Rapidly Compress Large Language Models without a Significant Loss of Quality The Yandex Research team, together with researchers from the Massachusetts Institute

We’re open-sourcing 352GB of Coral Reef pics (13 sites, 90k pics) from Indonesia under CC-BY-4.0 🌏🪸 3D photogrammetry data to accelerate research/conservation, no strings attached 🤗 🔵 Why? Coral reefs are so precious, beautiful, incredibly complex and threatened ecosystems.

Excited to share our work on Speculative Decoding Snowflake AI Research! 🚀 4x faster LLM inference for coding agents like OpenHands All Hands AI 💬 2.4x faster LLM inference for interactive chat 💻 Open-source via Arctic Inference as a plugin for vLLM 🧵

The LeRobot hackathon is now scheduled to happen in 44 different locations at the same time. Which city is missing: London (UK) - Cotono (Benin) - Toulouse, Paris & 2 in Lyon (France) - Antwerp (Belgium) - Santiago (Chile) - Isfahan (Iran) - Anchen, Berlin & Munich (Germany)

Idk who needs to see this, but the Unitree G1 with the back plate off:

🎉Factorio Learning Environment 0.2.0 released! 📖Details: jackhopkins.github.io/factorio-learn… New Features: - Multi-agent support - Reasoning models + MCP - Reflection & backtracking - Vision-augmented inputs and more frontier model results! The initial release of FLE was met with great

Someone slapped together a big dataset of password leaks today. Here's the distribution of pin numbers from a few times those got leaked a while back.

High quality image editing no longer needs closed models We release FLUX.1 Kontext [dev] - an open weights model for proprietary-level image editing performance. Runs on consumer chips. ✓ Open weights available ✓ Best in-class performance ✓ Self-serve commercial licensing

its outttt

New variant of attention by meta going beyond the standard bilinear form. It's changing the beta coef in scaling laws (which is a big deal) + there is a efficient triton implementation. Huge.

Another attention! Introducing Power Attention—a breakthrough in efficient attention mechanisms! This novel linear-cost attention layer features tunable state size, completely decoupled from model parameters. Why it stands out: ⚡ Blazing-fast GPU kernels with fused operations

Our latest CS336 Language Modeling from Scratch lectures are now available! View the entire playlist here: youtube.com/playlist?list=…

Interesting new Memory framework and paper released! MemOS claims to outperform competitors by treating memory as OS-like framework using a three-layer system: Interface, Operation, and Infrastructure. MemOS is an open-source library that claims to differentiates itself by: 1.

In physics there is an elegant method for computing the correlation functions called generating function. The idea is simple - instead of computing correlators one by one - you define a function of a parameter and compute the average of that new function. Individual correlators

Kimi team just trained a state of the art open source model 32B active parameter/1T total with 0 training instabilities, thanks to MuonClip, this is amazing

Tokenization has been the final barrier to truly end-to-end language models. We developed the H-Net: a hierarchical network that replaces tokenization with a dynamic chunking process directly inside the model, automatically discovering and operating over meaningful units of data

Why did Kimi.ai switched from closed source to open-source and released K2? - Reputation. If K2 is API only, it might have ended up like Grok-4 - clearly well-built yet still taking a lot of flak. - Work with the whole ecosystem. Less than 24 hours after release, the

Mistral released 2 open-source speech models: Voxtral (24B) & Voxtral Mini (3B). Both beat Whisper v3, GPT-4o-mini, and Scribe on ASR across 7+ languages, while supporting Q&A, summarization, and function-calling directly from voice. huggingface.co/mistralai

In our continued commitment to open-science, we are releasing the Voxtral Technical Report: arxiv.org/abs/2507.13264 The report covers details on pre-training, post-training, alignment and evaluations. We also present analysis on selecting the optimal model architecture, which

Today we are releasing FLUX.1 Krea [dev] - a new state-of-the-art open-weights FLUX model, built for photorealism. Developed in collaboration with KREA AI, this model is focused on images with unique aesthetics. No “AI look”, no blown-out highlights, just natural detail.