P(doom) on user profiles
Add P(doom) to user profiles: estimated probability of AI causing an existential catastrophe. Contributes to a community survey on AI risk perception.
Last updated · 29 releases tracked since November 2022
Hugging Face updates and release notes covering models, developer tools and training infrastructure on the Hugging Face Hub. Entries come from the Hugging Face Blog, and each one is dated and linked to the original post.
Hugging Face has shipped 2 updates so far in October 2026, mostly Product.
This month
2
-1 vs last month
Last 12 months
13
Busiest: March 2026
Most shipped
Developer Tools
5 releases
Newest first. Each entry links to the official page it came from.
Add P(doom) to user profiles: estimated probability of AI causing an existential catastrophe. Contributes to a community survey on AI risk perception.
New decision models added to llama.cpp via /v1/systemone endpoint, enabling probability-based option scoring with supported models like Kev-4B, OpenJev, and lev. Features include state-based input processing and API compatibility with System One format.
New Resources panel on job page shows live CPU, RAM, GPU, and VRAM usage during job execution.
LeRobot datasets now include episode previews with synchronized camera views and dataset metadata previews.
Hugging Face now available on Google Cloud Marketplace for unified billing with Google Cloud accounts.
Granular feature access control per resource group instead of organization-wide settings.
Hugging Face launched Storage Buckets on the Hub, adding mutable S3-like storage for checkpoints, logs, and other fast-moving ML artifacts.
Hugging Face released LeRobot v0.5.0 with humanoid support, new policies, faster dataset tooling, and EnvHub integration from the Hub.
Hugging Face introduced Modular Diffusers, giving developers composable building blocks for assembling and sharing diffusion pipelines.
Hugging Face announced that GGML and llama.cpp are joining HF, giving the local AI stack long-term resources while keeping the projects open-source and community-led.
Gradio's gr.HTML now supports custom templates, scoped CSS, and JavaScript interactivity, enabling single-file web app development with frontend, backend, and state management in Python.
Daggr is a new, open-source Python library for building AI workflows that connect Gradio apps, ML models, and custom functions. It automatically generates a visual canvas where you can inspect intermediate outputs, rerun individual steps, and manage state for complex pipelines,…
Hugging Face published an official explainer for Open Responses, a new open inference standard that extends the Responses API for agentic workloads.
BigCodeArena is a human-in-the-loop platform for evaluating code generation models through execution, allowing side-by-side comparisons with real-time code execution and community voting.
Hugging Face Transformers now integrates with SGLang as a backend, enabling high-performance inference for any Transformers-compatible model. Supports custom models from Hugging Face Hub and automatically falls back to Transformers when needed.
Comprehensive guide to knowledge distillation techniques, including its history, types, and real-world applications.
SmolVLM2 是一种轻量级视频语言模型,提供三种参数规模(22亿、5亿、2.56亿),支持MLX框架,适用于多种应用场景如视频理解、多图推理等。
This community article explains the role of profiling in agentic workflows, exploring how agents define their identity, behavior, environment, performance, and resource limits through frameworks like BDI and PEAS. It discusses how profiling serves as the foundation for intellige…
The blog post introduces the Algorithm of Thoughts (AoT), a heuristic-driven approach that outperforms traditional methods when applied to swarm algorithms like Particle Swarm Optimization (PSO) rather than Large Language Models (LLMs). It highlights key similarities with PPO al…
Streamlit app with Bokeh for interactive image similarity visualization using image embeddings
Introduces the Model Card Generator Interface, a tool for creating interactive HTML or static Markdown reports to document machine learning models with structured insights.
The Mixture of Agents Model (MAM) integrates specialized AI agents for front-end, back-end, database, DevOps, and project management into a unified system, streamlining software development through automation and efficiency.
Generative Password Manager (GPM) using neural networks to generate passwords from a master key, ensuring deterministic, high-entropy, and memorable password generation.
Indexify is an open-source data framework for building real-time, data-intensive applications using Hugging Face models.
A guide on outpainting using an inpainting model, detailing steps to generate and refine images with Hugging Face's tools.
Announces the Hallucinations Leaderboard, an open platform to evaluate large language models on hallucination-related benchmarks using frameworks like EleutherAI's Language Model Evaluation Harness. Includes tasks for closed-book QA, summarisation, reading comprehension, and mor…
Hugging Face 推出全面的推理(Inference)解决方案,包括推理组件、推理 API、推理 Endpoints 和推理 Spaces,支持开发者快速部署和生产级应用。
Announced integration of Intel OpenVINO with 🤗 Optimum Intel for accelerated model inference and quantization capabilities.
Releases per month over the last 12 months. Busiest month: March 2026, 3 releases.
Releases in the last 12 months by category and product.
The most recent Hugging Face update we track is “P(doom) on user profiles”, released on October 5, 2026.
Hugging Face shipped 2 updates in October 2026, including “P(doom) on user profiles”, “llama.cpp Decision Models Release”.
We tracked 13 updates from Hugging Face in the last 12 months, about 1 a month. The busiest month was March 2026, with 3 releases.
Hugging Face publishes release notes on Hugging Face Changelog, linked at the top of this page. Every entry here links back to the official page it came from.
Newsletter
This viewscreen turns the calendar into a recurring digest: the sharpest launches, model drops, and product moves delivered in one clear read.
The standout launches, model drops, and platform moves worth tracking.
One tight read instead of bouncing between company blogs and timelines.
Clear summaries you can scan fast and forward to the team when needed.
Monthly dispatch
Get a clean monthly dispatch covering the launches that mattered, not every minor product breadcrumb.
One email each month, tuned for signal over noise.