Search: “clem-delangue--x”

Introducing Decision 3.0: towards open multimodal foundation decision models ⚡ Open and state-of-the-art from 0.6B to 27B: #1 on both Jev Decision Index boards in our evaluation, on a new Pareto front

🧭 New model: Typed Decision, built on EmbeddingGemma 2 Give it a state + any schema of questions → typed, calibrated answers (choice · score · noul). No text generation, no output parsing. 76.4% on ty

We turned @OpenAI's math problems into open-source RL environments on @huggingface! Early and rough (the verifier only accepts the exact original formalization, so an equivalent proof can still score

We were able to fine tune an open weight model (Qwen3.5-9B) on proprietary internal data at @coinbase, and it out performed frontier models, at much lower cost and speed, for a specialized task (fraud

Finetuned Qwen3.5-4B into a decision model like Jev using @UnslothAI. - 4-bit QLoRA, bf16, single A100, 15 min. - Train/val loss flatten after 2 epochs so most of the learning happens fast. - Trained

SAM 3.1 + DINOv3 from @AIatMeta is a crazy combo 🔥 it can even hold onto surgical tools while a forearm sweeps over them. one leaves for 15 seconds, its near-identical twin stays, and it knows exactly

We distilled clef-flash into a 0.6B model and supporting ANE option. 1,000 support tickets triaged on a M5 pro in 57 s: team, urgency and refund for each, 48 ms per ticket. Within a few points of the

OpenAI math problems are now open source RL environments We’re releasing a broad range of new mathematical results produced by an internal frontier model. We’ve been consulting with the independent Ad

Training a robot to grasp an object means learning from many attempts, viewpoints and situations. One robot with three cameras at 20 fps produces 216,000 images per hour. Now that robotics is scaling,

TRL v1.15 is out, and it’s an absolute banger of a release for memory-efficient post-training. The main change: SFT, DPO, KTO, GRPO, RLOO and Distillation now use a fused LM head by default. Instead o

an open model just beat Jev outright on clinical decisions 🔥 pplx-decider v1.1 got 643 of 669 right vs 628 for Jev, at 42% less per decision and about the same speed. and you can download the weights

Today we are launching an agent arena for RL environments: The Open Env Arena. Your agents can work to define the best RL environments; the platform trains Qwen-3.8-27B on them, we evaluate the models

In October 2026, a Twitter user directed developers to build their own Reachy Mini using the Hugging Face Space “Reachy Mini – Yo It’s Your Boy” hosted by pollen‑robotics. The space provides an intera

On October 7, 2026, Clement Delangue noted that Microsoft’s Windows launch—highlighted by Satya Nadella and others—demonstrated the growing excitement around local AI solutions such as llamacpp. The e

Introducing River Recipes: our open-source initiative for turning promising research into vetted, reproducible training recipes. Our first recipe tackles text-to-SQL, where our trained open models bea

The largest agentic LLM Inference dataset has arrived! 206B tokens · 12,002 sessions · 1,186,582 LLM requests · 1,213,347 tool calls The traces include tool names, arguments, durations, and more. Expl

We show how you can use the River API to fine-tune open weight models to beat GPT-6 Astra and Claude Opus 5.5 on tasks you care about, with less than 1% of the cost! A few years from now most of the L

We’re open-sourcing pplx-embed-v2-late, multi-vector embeddings for text and images, 9B and 0.6B, in one shared embedding space. You can use these to index multimodal data with 9B, and query on device

We're releasing pplx-embed-v2-late, two late-interaction embedding models that retrieve text, images, and pages with a shared embedding space for cross-model querying. Both models achieve frontier per

When the open-source AI platform Hugging Face was hit by a cyberattack this July, it quickly became clear this was no ordinary breach because the hackers weren’t human: they were autonomous agents cre

Excited to share that I have (re)joined @huggingface as a Opensource ML engineer! Many incredible things to come around this fantastic technology -extremely grateful for the opportunity, I am very exc

Meet Mistral Large 4, aka Le Chonk. • 1T parameters, natively multimodal. 49B active. It is the best open weights model from US or Europe on aggregated benchmarks. • State-of-the-art on critical workl

Super excited about this and I know weights are coming (https://huggingface.co/mistralai/Mistral-Large-4.0-1T05-A52B) but just to be fair, you can't be the best open-weight model if you're not open-we

Introducing Agr, an open decision model by Command Code. · Agr (31B) and Agr-flash (360M) · 58.15 on Decision Index 0.2.1 · Focus on: tool calls & routing · TypeSafe SDK support Agr is best-in-class o

Let's go https://huggingface.co/reflection! Introducing Beam: a highly efficient agentic open model with 501B total parameters and 23B active. - Frontier reasoning efficiency - Advances the Western op

RL environments now have a home on the @huggingface Hub. Every RL framework has its own way to find environments. Custom hubs, runtime registries, GitHub lists with a custom loader. If you publish for

Thanks for sharing! We open-sourced code and models. 🤗Huggingface Paper: https://huggingface.co/papers/2609.38597 💻 Code: https://github.com/nv-tlabs/PixelUMM 🤗 Model: https://huggingface.co/nvidia/Pi

The new v0.6.0 release packs a lot of good stuff: - Clef support (text + vision) - High-quality support for Qwen3.8-Flash-Next - Massive performance improvement with Metal - New `llama_batch_ext` API

We turned Claude Code, Codex, Hermes, Pi, @opencode and other coding harnesses into RL environments. No changes to the harnesses, no changes to the training code. Any open model, any task set, fully o

A tweet by Kristof (@CoastalFuturist) titled “How it feels to ask the Chinese ai models for help on something Claude won’t agree to” was posted at 1:30 AM on Oct 4, 2026. The post garnered 616.3 k vie

HuggingFace just closed a major gap in harness engineering! (the ultimate guide to multi-harness RL) the same open-weight model can perform well inside one harness, then lose accuracy or produce inval

Microsoft CEO Satya Nadella wrote that a company renting an AI model “essentially pays for intelligence twice”: Once in money, and again in the proprietary knowledge it has to reveal to make the model

I just made the most expressive real-time motion harness for the Reachy Mini. It was created by: > synthesizing a massive motion dataset using Astra (over 10000 samples) > fine-tuning Qwen 3.5 4B usin

Agentic RL is hard at every layer: environments, training infra, and algorithms. So we’re open-sourcing all of it. For environments, that means both the dataset and the scalable synthesis + quality-co

Decision models in llama.cpp are now available The `/v1/systemone` endpoint is available in the latest llama builds. Use it to do Jev-style inference locally, efficiently and privately. Multiple open

It is here now ! Enjoy our 4B model 🐸 https://huggingface.co/microsoft/FrogNano-4B-2609 4B parameters. ZERO distillation. 61.5% on SWE-bench Verified 🤯 Meet FrogNano 🐸: Qwen3.5-4B post-trained purely

SETA got accepted to NeurIPS 2026! 🎉 We released the full training scripts for DeepSeek-V4-Flash, GLM-5.2 (LoRA), Inkling-Small, and Qwen3.8-27B, with both GRPO and PPO, alongside 4,500+ open-source,

Thank you @huggingface and @pollenrobotics for lending us a Microduck preview unit for DevDay! 🤗 Just a couple hours with Astra to control the duck and add voice, vision… I haven’t even shown visual r

The same model, with the same weights, scores 62% in one agent harness and 33% in another. @adithya_s_k and the @huggingface team just released the ultimate guide to multi-harness RL, and it's one of

This is beautiful work @huggingface! 🤗 The same model, with the same weights, scores 62% in one agent harness and 33% in another. @adithya_s_k and the @huggingface team just released the ultimate guid

Today we're unveiling Trillium Labs @trillium_labs, a new non-profit to foster the open science of frontier AI. We're building open post-training recipes and will expand into open infra to study RSI,

PerplexityAI has released a new language‑model variant on Hugging Face titled **pplx-decider‑v1‑27b**, described as a “Jev alternative.” The announcement was posted by Victor M (victormustar) at 8:05

babe, wake up. Hugging Face just dropped a new banger article about RL training coding agents on different harnesses. 1/ Excited to release The ultimate guide to multi-harness RL The same model behave

introducing 𝚌𝚕𝚎𝚏: our first models trained by @cloudflare's workers ai team. today, we're releasing two fast and accurate decision models that top the benchmarks for quality and latency. use them host

pplx-decider-v1-27b is fine-tuned from Qwen3.8-27B with a 250k context window. View the weights on Hugging Face: https://huggingface.co/perplexity-ai/pplx-decider-v1-27b Get started with the Perplexit

🪽DFlash draft models are available for Ornith-1.5 9B, 35B-A3B, and 397B. DFlash and Ornith 1.5 make a dynamic duo: DFlash performs as the draft model, accelerating inference by drafting multiple token