model releases

This post from Anthropic's X account shares a Wordle game result, specifically puzzle #1,772 solved in 3 attempts. The colored squares indicate the progression of guesses, with yellow squares showing

2 bit DeepSeek v4 Flash inference. Experts w1/3: IQ2_XXS, w2: Q2_K, all the rest left mostly F16/F32 Final GGUF: 86.18 GiB Runs on a MacBook M3 max with CPU (No Metal backend test as it will crash my

DeepSeek released V4 Pro (1.6T-A49B) and Flash (284B-A13B) models in both base and instruct variants, with optimizations enabling them to run on Huawei Ascend chips. These releases represent updates t

Alpha Eval: Agents Making Evals as a Multi-Player Game This diagram and blurb is largely a research riff with Claude on building data generation systems to get closer to the holy grail of self-improvi

Simon Willison inquires about running DeepSeek-V4-Flash on Mac hardware, specifically asking about feasibility across different RAM configurations from 512GB down to smaller amounts. This reflects dis

Qwen's latest model improvements focus on achieving more consistent and uniform performance quality across different artistic styles and aesthetic domains, reducing the variability in output quality t

btw we are cooking something with @hhua_ (not final yet but keep calendar open after ICML in Seoul) 🚀 DeepSeek-V4 Preview is officially live & open-sourced! Welcome to the era of cost-effective 1M con

Comprehensive Multilingual Text Rendering： Better glyph accuracy, more consistent typography, and cleaner layouts even in complex compositions. It handles mixed-language scenarios more gracefully as w

DeepSeek V4, a large-scale mixture-of-experts (MoE) model, is now available on Hugging Face with support from Novita offering 1 million token context window. The model is offered in two variants—Pro a

🔥DeepSeek-V4-Pro API is 75% OFF until May 5th, 2026, 15:59 (UTC Time)! Don't miss out on this massive discount. 🛠️Integration Updates: 🔹Claude Code: Set model to deepseek-v4-pro[1m] to unlock 1M conte

gpt-5.5 is now available in the ml-intern! this means it gets access to the whole @huggingface infra: buckets, jobs, repos etc for doing ai research at scale giving it a spin now to see if i'm even cl

GPT-5.5 prompting guide Now that GPT-5.5 is available in the API, OpenAI have released a wealth of useful tips on how best to prompt the new model. Here's a neat trick they recommend for applications

grok imagine is on another level now. the new model is unreal. go try it. Grok Imagine now has dramatically improved lip sync and sharper audio quality on all image-to-video generations. Dialogue trac

Hamming's talk is so important that I reproduced it on my site. It's one of the only things on my site written by someone else. https://paulgraham.com/hamming.html A mathematician who shared an office

I crossed 100K followers on X and walked the red carpet at the TIME 100 gala as a TIME 100 AI honoree. What a week. Claude and I have decided the only logical next step is 100 lotto tickets. If I win

Kimi K2.6 from @Kimi_Moonshot is now available on @FireworksAI_HQ Training Platform across the Managed and Training API workflows. Try SFT, DPO, RL with smart defaults or your own custom loss function

Local models do seem likely to create an explosion of new use cases. Local compute >> cloud compute. This is where we are right now. And i’m not gonna lie it feels pretty magical 🧚‍♀️ Qwen3.6 27B runn

🚀Meet Carnice-V2-27b🚀 → Carnice is a 27 billion parameter model capable of beating models 10x the size in Hermes-agent, fully open-source and built on top of Qwen3.6-27B →Build to fit on Consumer GPU

OpenAI just released HealthBench Professional on Hugging Face A medical evaluation benchmark designed to improve AI assistants for clinicians, featuring physician-curated conversations and rubric-base

Our GB300 cluster went down yesterday, just as Deepseek released 😱 We were 😥 but @CoreWeave came through to contribute to the Open Source. They scrambled in the compute crisis, finding 2 spare dev rac

Since GPT-5.4, we’ve unified Codex and the main model into a single system, so there’s no separate coding line anymore. GPT-5.5 takes this further, with strong gains in agentic coding, computer use, a

Qwen-Image-2.0-Pro is now live 🚀🚀 We’ve pushed image quality, multilingual text rendering, and instruction following to a new level, while making performance much more consistent across styles.🌅🌃 Rank

SDR to HDR from ComfyUI, I've trained a LoRA over Qwen Edit 2011 based on the principle used in https://hdr-lumivid.github.io/ A research I've had the previlige to work on with @noamiKenKorem for vide

Spot on 🎯 We don't need frontier intelligence to automate searches and sending emails - We don't need trillion parameter models to be able to summarize articles or technical documents - We don't need

This is one you should watch on a big TV. One of the most incredible corporate videos I ever have seen and I was with the GoPro video team the night it released its Hero 2 video that brought that comp

We built HF for AI builders collaboration, fun to see it's increasingly becoming the place for agent collaboration! This morning I'm sending my ml-intern to participate in the @OpenAI Parameter Golf c

We've got all the models here: https://dell.huggingface.co/authenticated/models Kimi K2.5, Mistral, Cohere, Arcee AI Trinity Large, Google Gemma, Meta/Llama, Qwen, Nvidia Nemotron, Grok, GPT OSS, Deep

What if instead of building one giant AI, we evolved a coordinator to orchestrate a diverse team of specialized AIs? 🐟 Excited to share our new paper: “TRINITY: An Evolved LLM Coordinator”, published

@scottjla on Twitter in reply to my pelican riding a bicycle benchmark: I feel like we need to stack these tests now I checked to confirm that the model (ChatGPT Images 2.0) added the 'WHY ARE YOU LIK

The Qwen 3.6-27B and 35B models are designed as open-weight models that developers can run locally on their own hardware without requiring cloud services. Alibaba released a separate cloud-only produc

This is a Wordle game result from Anthropic's X (Twitter) account showing they solved puzzle 1,770 in three attempts, with the final word being a five-letter combination indicated by the green squares

Google Cloud Next ‘26 took place this week in Las Vegas, and the energy was incredible as we welcomed over 32,000 leaders, developers, and partners to explore the Agentic Era with us. Across three key

4⃣4⃣4⃣4⃣ Introducing DeepSeek V4 Pro, a long-context model with hybrid attention, three reasoning modes, and SOTA coding performance. AI natives can now use DeepSeek V4 Pro on Together AI and benefit

DeepSeek-V4 rapidly gained over 500 likes within 28 minutes on Hugging Face, demonstrating exceptional user engagement and positioning it as a strong contender to become the fastest model to reach #1

arXiv:2601.18672v3 Announce Type: replace Abstract: Kolmogorov-Arnold Networks (KANs) have recently demonstrated promising potential in scientific machine learning, partly due to their capacity for gr

arXiv:2604.21411v1 Announce Type: new Abstract: Standard physics-informed neural networks (PINNs) struggle to simulate highly oscillatory Helmholtz solutions in heterogeneous media because pointwise m

arXiv:2604.21688v1 Announce Type: cross Abstract: The IC3 algorithm represents the state-of-the-art (SOTA) hardware model checking technique, owing to its robust performance and scalability. A signifi

A mathematician who shared an office with Claude Shannon at Bell Labs gave one lecture in 1986 that explains why some people win Nobel Prizes and other equally smart people spend their whole lives doi

arXiv:2604.21579v1 Announce Type: cross Abstract: LLM-based automated program repair (APR) techniques have shown promising results in reducing debugging costs. However, prior results can be affected b

arXiv:2604.21623v1 Announce Type: cross Abstract: The proliferation of Internet of Things (IoT) devices has significantly expanded attack surfaces, making IoT ecosystems particularly susceptible to so

arXiv:2604.20915v1 Announce Type: cross Abstract: Transformers suffer from a high computational cost that grows with sequence length for self-attention, making inference in long streams prohibited by

arXiv:2604.20932v1 Announce Type: cross Abstract: Retrieval-augmented generation (RAG) systems are increasingly deployed in sensitive domains such as healthcare and law, where they rely on private, do

arXiv:2604.20846v1 Announce Type: cross Abstract: Next point-of-interest (POI) recommendation requires modeling user mobility as a spatiotemporal sequence, where different behavioral factors may evolv

arXiv:2604.21725v1 Announce Type: cross Abstract: LLM agents increasingly operate in open-ended environments spanning hundreds of sequential episodes, yet they remain largely stateless: each task is s

arXiv:2604.20996v1 Announce Type: new Abstract: How can language learning systems be developed for languages that lack sufficient training resources? This challenge is increasingly faced by developers

arXiv:2601.18491v2 Announce Type: replace Abstract: The rise of AI agents introduces complex safety and security challenges arising from autonomous tool use and environmental interactions. Current gua