✷ The archive · 11 labs · 3,060 dispatches
The labs
No more opening a dozen official blogs every morning. First-hand releases from OpenAI, Anthropic, DeepMind and the rest, each with its substance pulled out.
OpenAI and California State University System Deployment of ChatGPT Edu
OpenAI and the California State University (CSU) system have launched the largest deployment of ChatGPT to date, providing ChatGPT Edu access to over 520,000 students and faculty across 23 campuses.
DABStep: Data Agent Benchmark for Multi-step Reasoning
Hugging Face and Adyen have introduced DABStep, a benchmark of over 450 real-world data analysis tasks that reveals current state-of-the-art AI agents achieve only 16% accuracy on complex reasoning tasks.
pi0 and pi0-FAST: Vision-Language-Action Models for General Robot Control
Hugging Face has integrated pi0 and pi0-FAST, generalist Vision-Language-Action (VLA) models developed by Physical Intelligence, into the LeRobot repository to enable versatile robot control across diverse embodiments.
Using ChatGPT for Nail Art Design
Licensed nail technician Tabytha Scott uses ChatGPT as a creative partner to narrow down infinite design ideas into executable nail art concepts.
Hugging Face Open-source DeepResearch
Hugging Face has open-sourced a framework to reproduce OpenAI's Deep Research capabilities, achieving a 55.15% score on the GAIA benchmark using a code-native agentic approach.
Using ChatGPT for Specialized Fishing Guidance
OpenAI highlights how a professional angler uses ChatGPT to develop targeted fishing plans and accelerate the learning curve for beginners in the sport.
Building a Custom Math Tutor with ChatGPT GPTs
Phil Birchenall used a custom GPT to create a personalized, dog-themed math tutor that helped his daughter improve her math skills and pass her SATs exams.
OpenAI Deep Research for Complex Industry Trends
OpenAI has introduced deep research, a tool designed to automate the analysis of complex industry trends to increase the personal capacity of professional researchers.
OpenAI Deep Research
OpenAI has launched deep research, an agentic capability in ChatGPT powered by a version of the o3 model that autonomously conducts multi-step web research to produce comprehensive, analyst-level reports.
Open-R1 Update #1: Replicating DeepSeek-R1 Training and Synthetic Data
Hugging Face provides a first progress update on the Open-R1 project, detailing the reproduction of DeepSeek-R1 evaluation scores, the integration of GRPO into TRL, and strategies for scaling synthetic reasoning data generation.
OpenAI Disrupts AI-Assisted Romance-Baiting and Pig Butchering Scams
OpenAI has banned a cluster of ChatGPT accounts used by a Cambodia-based network to automate translation and persona generation for romance and investment scams.
OpenAI Disrupts 'Sponsored Discontent' Influence Operation
OpenAI has banned accounts linked to a Chinese influence operation, dubbed 'Sponsored Discontent,' which used ChatGPT to generate critical content about the US and a Chinese dissident, eventually planting long-form articles in Latin American mainstream media.
OpenAI Disrupts DPRK-Affiliated Cyber Threat Actors Using AI
OpenAI has banned accounts linked to DPRK-affiliated threat actors, including VELVET CHOLLIMA and STARDUST CHOLLIMA, who used AI for coding assistance, debugging, and social engineering research.
OpenAI Disrupts Iranian Influence Nexus Using ChatGPT
OpenAI has banned five ChatGPT accounts used by Iranian-linked influence operations STORM-2035 and IUVM to generate pro-Iran and pro-Hamas content across multiple languages and platforms.
OpenAI Disrupts Covert Influence Operation Targeting Ghana Election
OpenAI banned a cluster of ChatGPT accounts used by a commercial entity called DigitSol to generate pro-Bawumia and anti-Mahama content for a covert influence operation during the Ghanaian presidential election.
OpenAI Disrupts AI-Assisted Task Scam Network
OpenAI has banned a cluster of ChatGPT accounts used by a Cambodia-based network to facilitate 'task scams' through AI-powered translation and social engineering.
OpenAI Disrupts AI-Assisted Deceptive Employment Scheme
OpenAI banned dozens of accounts used in a deceptive employment scheme that leveraged AI to fabricate job applications, conduct interviews, and perform work tasks, aligning with TTPs attributed to North Korean state efforts.
OpenAI Operation Peer Review: Disrupting AI-Assisted Surveillance Planning
OpenAI banned a cluster of ChatGPT accounts likely originating in China that used AI to assist in researching political actors, analyzing protest documents, and developing surveillance tools.
OpenAI o3-mini System Card
OpenAI released the o3-mini system card, detailing how large-scale reinforcement learning and chain-of-thought reasoning enable the model to achieve state-of-the-art safety performance and a Medium risk classification under the Preparedness Framework.
OpenAI o3-mini release notes / what's new
OpenAI has released o3-mini, a cost-efficient reasoning model optimized for STEM, offering performance comparable to o1 in math, coding, and science with lower latency and expanded developer features.
Mini-R1: Reproducing DeepSeek-R1 Reasoning via GRPO and the Countdown Game
Hugging Face demonstrates how to reproduce the DeepSeek-R1 "aha moment" of self-correction and reasoning by training a Qwen2.5-3B model using Group Relative Policy Optimization (GRPO) on the Countdown Game.
Hugging Face AI Tools for Art Newsletter Issue 1
Hugging Face has launched a monthly newsletter detailing the state of open-source creative AI, highlighting 2024's shift to Diffusion Transformers and the emergence of high-quality open video and audio models.
Mistral Small 3 Release Notes
Mistral AI has released Mistral Small 3, a latency-optimized 24B-parameter model under the Apache 2.0 license that rivals the performance of models three times its size.
OpenAI and U.S. National Laboratories Partnership
OpenAI has partnered with the U.S. National Laboratories to deploy reasoning models on the Venado supercomputer to accelerate scientific research and strengthen national security.
How to Deploy and Fine‑Tune DeepSeek R1 Models on AWS
Hugging Face shows how to deploy and fine‑tune DeepSeek R1 and its distilled variants on AWS using Hugging Face Inference Endpoints, Amazon Bedrock Marketplace, Amazon SageMaker AI (GPU and Neuron instances), and EC2 Neuron with the Hugging Face Neuron Deep Learning AMI.
Qwen2.5-Max Release Notes
Qwen has released Qwen2.5-Max, a large-scale Mixture-of-Experts (MoE) model pretrained on over 20 trillion tokens and available via API and Qwen Chat.
Open-R1: A Fully Open Reproduction of DeepSeek-R1
Hugging Face has launched the Open-R1 project to systematically reconstruct the data and training pipeline of DeepSeek-R1, aiming to provide the open-source community with the missing datasets and code for reasoning models.
Hugging Face Inference Providers Integration
Hugging Face has integrated four serverless inference providers—fal, Replicate, SambaNova, and Together AI—directly into the Hub to provide unified, model-centric serverless inference.
State of Open Video Generation Models in Diffusers
Hugging Face provides a comprehensive overview of open video generation models and introduces a suite of Diffusers optimizations that can reduce VRAM requirements for models like HunyuanVideo from 60GB to approximately 6.5GB.
Qwen2.5-1M Release: Open‑Source 7B and 14B Models with 1M‑Token Context and vLLM‑Based Inference Framework
Qwen releases open‑source Qwen2.5‑7B‑Instruct‑1M and Qwen2.5‑14B‑Instruct‑1M models that support up to 1 million token contexts, accompanied by an optimized vLLM‑based inference framework that delivers 3×–7× faster prefill and retains short‑task performance comparable to GPT‑4o‑mini.
Qwen2.5-VL Release Notes
Qwen has released Qwen2.5-VL, a flagship vision-language model available in 3B, 7B, and 72B sizes that introduces advanced visual agent capabilities, long-video comprehension, and structured document parsing.
smolagents Vision Support Update
Hugging Face has added native vision support to smolagents, enabling the use of Vision Language Models (VLMs) in agentic pipelines for tasks like autonomous web browsing.
OpenAI Operator Research Preview
OpenAI has introduced Operator, a research preview of an AI agent powered by the Computer-Using Agent (CUA) model that can independently navigate a web browser to perform tasks like filling forms and ordering groceries.
OpenAI Computer-Using Agent (CUA) and Operator Research Preview
OpenAI has introduced Computer-Using Agent (CUA), a model capable of interacting with GUIs via a universal interface of screen, mouse, and keyboard to perform digital tasks, now available in research preview via Operator.
OpenAI Operator System Card
OpenAI has introduced Operator, a research preview of a Computer-Using Agent (CUA) model that interacts with GUIs to perform tasks on a user's behalf, supported by a multi-layered safety framework to mitigate risks like prompt injection and model mistakes.
NVIDIA KVPress: Toolkit for KV Cache Compression in Long-Context LLMs
NVIDIA has released KVPress, a Python toolkit that implements state-of-the-art KV cache compression techniques to reduce the memory footprint and increase decoding speed of long-context Large Language Models.
SmolVLM 256M and 500M Release Notes
Hugging Face introduces SmolVLM-256M and SmolVLM-500M, delivering highly efficient Vision Language Models that maintain strong multimodal performance in a significantly reduced parameter footprint.
Bertelsmann and OpenAI Partnership for Enterprise AI Integration
Bertelsmann is integrating OpenAI technology and ChatGPT Enterprise across its global media, services, and education brands to enhance creativity, productivity, and product development at scale.
Trading Inference-Time Compute for Adversarial Robustness
OpenAI research indicates that increasing inference-time compute in reasoning models like o1-preview and o1-mini improves their robustness against multiple types of adversarial attacks.
Hugging Face and FriendliAI Partnership for Model Deployment
Hugging Face has integrated FriendliAI's inference infrastructure into the Hugging Face Hub, enabling one-click deployment of generative AI models to high-performance endpoints.
OpenAI Stargate Infrastructure Initiative
OpenAI is seeking partnerships with US-based data center infrastructure firms to develop the physical capacity required for its next-generation AI systems.
OpenAI Announces The Stargate Project
OpenAI and its partners have launched The Stargate Project, a $500 billion initiative to build massive AI infrastructure in the United States to support the development of AGI.
Hugging Face Organization Blog Articles Feature
Hugging Face now allows organizations subscribed to Enterprise Hub to publish blog articles directly to their organization profiles.
Qwen Global-Batch Load Balance for MoE LLM Training
Qwen introduces a global-batch load balancing loss that improves MoE model performance and enables expert specialization by calculating balance across the entire global batch rather than individual micro-batches.
DeepSeek-R1 Release – Open-Source Model Matching OpenAI o1 Performance
DeepSeek announced the open-source DeepSeek‑R1 model, which matches OpenAI o1 on math, code, and reasoning tasks and is released under an MIT license with free commercial use.
Mistral AI Announces AFP Newswire Integration for Le Chat
Mistral AI announced a partnership with Agence France-Presse that integrates AFP's multilingual newswire into its Le Chat assistant, boosting factual accuracy and coverage for users.
Hugging Face Text Generation Inference (TGI) Multi-Backend Support
Hugging Face has introduced a multi-backend architecture for Text Generation Inference (TGI), allowing users to utilize a unified frontend to deploy LLMs via various execution engines like TensorRT-LLM and vLLM.
Hugging Face Transformers timm Integration
Hugging Face has introduced the TimmWrapper, allowing any model from the PyTorch Image Models (timm) library to be used seamlessly within the transformers ecosystem for inference, quantization, and fine-tuning.
OpenAI News Industry Partnerships and Axios Expansion
OpenAI has announced a new content partnership with Axios and expanded funding for local news, while detailing how nearly 20 media organizations are integrating AI to streamline production, enhance user engagement, and optimize business operations.
Train 400x faster Static Embedding Models with Sentence Transformers
Hugging Face introduces a method to train static embedding models that run 100x–400x faster on CPU while retaining at least 85% of the quality of models like all‑mpnet‑base‑v2, releasing two models (static‑retrieval‑mrl‑en‑v1 and static‑similarity‑mrl‑multilingual‑v1) with training scripts and evaluation results.