Search: “clem-delangue--x”

Congrats to @poolsideai on Laguna XS.2 release. Day 0 support in @vllm_project. FP8, INT4, and NVFP4 quantized checkpoints created with LLM Compressor also ready today. 33B params, 3B active. Apache 2

.@huggingface unveiled ml-intern – an open-source agent that automates the gritty post-training loop: - reading papers - tracing citations - curating datasets - running experiments - and iterating lik

I post-trained Qwen3-Coder to fix bugs using an actual debugger. The result: Solve rate: 70% → 89% Median turns to fix: 46 → 19 (-59%) Instead of just reading code or print-debugging, it: - reasons fr

'If they ever tell my story let them say that I walked with giants'--Troy I am humbled&excited the model we released last week is trending #2 on @huggingface, between giant models such as DeepSeek,Qwe

I’ve been saying this for over a year now: Frontier models are fantastic, but the real future is frontier-level models (in every way) running locally on your own hardware. I think this is 18-24 months

I've been trying to make transformers more agent-friendly: agentic CLI, a skill, doc rewrites, canonical examples. It felt a bit like shooting in the dark: hard to measure progress, and hard to ensure

Just received new reach minis for the Miami office! This is the first robot that goes out to people with agentic software meaning anyone can build any apps with their agents so excited to see what peo

LOCAL AI MODELS ARE CATCHING UP TO FRONTIER MODELS WAY FASTER THAN ANYONE EXPECTED this guy ran qwen 3.6 27B locally on a base macbook pro M4 with 24GB of memory quantized and stripped of safety guard

new in ml-intern: you can now actually see what's going on inside added native metric logging + trackio integration. every training run the agent kicks off now has live curves you can watch in real ti

Pi + local models are definitely really cool! Short demo to clean up my Desktop: > terminal 1: llama-server -hf unsloth/Qwen3.5-9B-GGUF:UD-Q4_K_XL > terminal 2: simply type 'pi' and start talking to i

Today @poolsideai is releasing Laguna M.1 & Laguna XS.2, our latest generation models and first public models We started Poolside because we believed that to build truly capable coding agents, you nee

Today we’re releasing Laguna XS.2, Poolside’s first open-weight model. It’s a 33B total / 3B active MoE model built for agentic coding and long-horizon tasks. Trained fully in-house on our own stack.

Today we’re shipping Laguna M.1 and Laguna XS.2 – our first public models. We’re also shipping our agent harness and a preview product experience. Both models were trained from scratch on our own stac

This post provides links to resources for SWE-Chat, including the arXiv paper (2604.20779), an interactive website (swe-chat.com), and a dataset hosted on Hugging Face. SWE-Chat appears to be a conver

A completely local agent that lives right inside your browser. Powered by Gemma 4 E2B and WebGPU, it uses native tool calling to: 🔍 Search browsing history 📄 Read and summarize pages 🔗 Manage tabs 100

going to buy 2 rtx 6000 just because of the capabilities of local models becoming great! no more outsourcing of research to the api Xiaomi MiMo-V2.5 is now officially open-sourced！ MIT License, suppor

Lately I've been having fun with running coding agents fully locally. The setup I landed on is: - Pi agent - Gemma 4 26B A4B - Server of choice: LM Studio/Ollama/llama.cpp I wrote a step-by-step guide

Llamacpp is a tool for running large language models locally on personal computers, offering free access without relying on cloud services. The technology emphasizes speed, security, and computational

ml-intern from @huggingface is such a cool repository, an 🤖 agentic harness that allows you to wield the larger🤗ecosystem to run all the tasks a ml researcher would. it's all one can ask for as a side

ml-intern now notifies you when it needs input! Amazing for those 12h+ autonomous LLM training runs The ML Intern can now ping you on Slack when it's finished training models, generating datasets, or

Ornstein-Hermes-27B-MLX live https://huggingface.co/GestaltLabs/Ornstein-Hermes-3.6-27b-MLX-16bit https://huggingface.co/GestaltLabs/Ornstein-Hermes-3.6-27b-MLX-8bit https://huggingface.co/GestaltLabs

ML Intern, a feature or tool, now supports Slack notifications to alert users when machine learning tasks complete, including model training, dataset generation, and ablation study analysis. This inte

Clem Delangue, CEO of Hugging Face, discusses best practices for releasing model weights in the open-source AI community and advocates for other AI organizations to follow suit in making their models

This post highlights the three most popular models on Hugging Face during a given week, featuring DeepSeek AI, OpenAI, and Alibaba's Qwen models. The post was shared by Clem Delangue, CEO of Hugging F

Clem Delangue argues that the United States should increase investment and policy support for open source AI models to maintain competitive advantage and reduce dependence on proprietary systems contr

@badlogicgames Yup: https://x.com/antirez/status/2048344770234249588 The GGUF tool calling template is wrong but I'm uploading a new GGUF file. Otherwise there is the right template file in one of the

DeepSeek v4 Flash with *local inference* after 24h of playing with that: even with the 2 bit selective quantization GGUF, iti is the FIRST time I feel I have a frontier model running on my computer. T

extremely happy that we are in q2 2026, and engineers i look up to are plowing a path for the local model future. yes, we are still beholden to some lab publishing weights. but i take that over compan

🚨GUYS QWEN3.6 27B/35B @unslothai MLX QUANTS @Brooooook_lyn cooked 🔥 Apple Silicon users, the exact models y’all have been asking for LANDED. Unsloth mixed-precision quants in native MLX format. >Fast.

HealthBench Professional, our new evaluation of clinician chat tasks, is now available on HuggingFace for easy access. Each example was written, reviewed, and adjudicated by three or more physicians.

Clem Delangue, co-founder of Hugging Face, shared a positive comment about the machine learning internship experience at Hugging Face, suggesting the internship program is engaging and enjoyable for p

A quantized version of DeepSeek-V4-Flash has been released on Hugging Face by the MLX community in 2-bit format with dynamic quantization, making the model more efficient for inference on resource-con

OpenAI released a new dataset on Hugging Face Hub designed to improve ChatGPT's performance in clinical settings, addressing the specific needs and workflows of healthcare professionals. The dataset s

Opened a llama.cpp discussion about whether custom GBNF grammars can compose with tool calls in llama-server. Right now tools work alone, grammar works alone, but tools+grammar doesn't. If you use lla

pay attention anon. this is what local ai actually feels like in 2026. qwen 3.6 27b dense just knocked down the second test in my single file agentic benchmark series. on 1x 3090. mandelbrot fractal e

The community can now download pre-quantized weights from MLX community repo on HF thanks to @LambdaAPI Model collection: https://huggingface.co/collections/mlx-community/deepseek-v4 DeepSeek-V4-Flash

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 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

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

This post describes a simple two-line process for uploading a personal knowledge base or 'second brain' to a private Hugging Face bucket using the Hugging Face CLI, with the suggestion that an AI agen

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

A new Dflash drafting model based on Qwen 3.6 with 27 billion parameters has been released on Hugging Face, available at z-lab/Qwen3.6-27B-DFlash. This model likely implements speculative decoding or

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

This post discusses the need for built-in speech-to-text (STT) and text-to-speech (TTS) capabilities in a product or platform (likely referring to Hugging Face's platform based on the mention of Clem

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