mmBERT: ModernBERT goes Multilingual
Hugging Face introduces mmBERT, a massively multilingual encoder model trained on 3T+ tokens across 1,800+ languages that outperforms XLM-R in both performance and inference speed.
OpenAI People-First AI Fund
OpenAI has opened applications for a $50 million fund providing unrestricted grants to U.S.-based nonprofits to advance AI literacy, community innovation, and economic opportunity.
OpenAI Research: Why Language Models Hallucinate
OpenAI argues that language models hallucinate because standard training and evaluation procedures reward guessing over acknowledging uncertainty, rather than due to a mysterious glitch.
OpenAI for Greece partnership launches education pilot and AI accelerator
OpenAI announced the ‘OpenAI for Greece’ partnership, launching a ChatGPT Edu pilot in secondary schools and a Greek AI Accelerator to boost local AI startups, marking a major step in embedding advanced AI across Greece’s education and innovation sectors.
OpenAI Expanding Economic Opportunity with AI Initiatives
OpenAI has announced the OpenAI Jobs Platform and OpenAI Certifications to help workers gain AI fluency and connect them with employers, with a goal of certifying 10 million Americans by 2030.
EmbeddingGemma 300M release notes
Google released EmbeddingGemma, a 308M‑parameter multilingual embedding model that tops the MTEB benchmark for sub‑500M models and is optimized for on‑device use.
SandboxAQ SAIR Dataset Release
SandboxAQ has released SAIR, the largest dataset of co-folded 3D protein-ligand structures paired with experimental IC50 labels, providing over 5 million AI-generated structures to accelerate AI-powered drug discovery.
OpenAI Acquires Statsig and Appoints Vijaye Raji as CTO of Applications
OpenAI has acquired the experimentation platform Statsig and appointed its founder and CEO Vijaye Raji as CTO of Applications to lead product engineering for ChatGPT and Codex.
OpenAI Updates ChatGPT Safety and Well-Being Initiatives
OpenAI is implementing a 120-day initiative to improve ChatGPT's response to mental health crises and introduce parental controls for teens, guided by a network of medical experts.
Hugging Face ZeroGPU Ahead-of-Time Compilation Guide
Hugging Face introduces ahead-of-time (AoT) compilation for ZeroGPU Spaces, enabling speedups of 1.3x to 1.8x for generative models like Flux, Wan, and LTX by eliminating just-in-time compilation overhead.
What word2vec learns: A closed-form theory of feature learning
BAIR researchers have developed a predictive theory proving that word2vec learning reduces to unweighted least-squares matrix factorization, where learned representations are effectively the result of PCA on a specific target matrix.
OpenAI gpt-realtime and Realtime API General Availability
OpenAI has released gpt-realtime, a high-performance speech-to-speech model, and made the Realtime API generally available with support for image inputs, remote MCP servers, and SIP phone calling.
OpenAI Supporting Nonprofit and Community Innovation Fund
OpenAI is opening applications for a $50 million fund to provide unrestricted grants to U.S.-based 501(c)(3) nonprofits and community organizations to foster AI-driven innovation for the public good.
OpenAI Collective Alignment and Model Spec Updates
OpenAI has introduced a collective alignment research effort to integrate public preferences from over 1,000 global participants into its Model Spec to ensure AI default behaviors reflect diverse human values.
OpenAI and Anthropic Joint Safety Evaluation Findings
OpenAI and Anthropic released joint safety evaluation results showing Claude 4 models excel at instruction hierarchy while OpenAI's reasoning models lead on jailbreak resistance, with both labs highlighting the value of cross‑lab collaboration for AI alignment.
OpenAI Mental Health and Crisis Intervention Safeguards
OpenAI has detailed its layered safety stack for managing mental health crises in ChatGPT, including the introduction of GPT-5's improved response accuracy and plans for direct emergency service integration.
OpenAI Learning Accelerator India Launch
OpenAI has launched the Learning Accelerator in India to provide AI research, training, and ChatGPT licenses to millions of educators and students through strategic institutional partnerships.
OpenAI and Retro Biosciences: Accelerating Life Sciences with GPT-4b micro
OpenAI and Retro Biosciences developed GPT-4b micro, a specialized protein engineering model that re-engineered Yamanaka factors to achieve over 50-fold higher stem cell reprogramming expression than wild-type controls.
Blue J Scaling Tax Research with GPT-4.1
Blue J leverages GPT-4.1 and a proprietary RAG system to automate complex tax research across the US, Canada, and the UK, reducing the disagree rate to fewer than 1 in 700 responses.
NVIDIA Nemotron Post-Training Dataset v2 and Nemotron Nano 2 9B Release
NVIDIA has released a 6-million sample multilingual reasoning dataset and the Nemotron Nano 2 9B model, which utilizes a hybrid Transformer-Mamba architecture to optimize reasoning costs and throughput.
MIXI ChatGPT Enterprise Deployment
MIXI deployed ChatGPT Enterprise across its organization in 45 days, reducing work hours by over 90% in some projects and enabling employees to create over 1,600 custom GPTs.
Generate Images with Claude and Hugging Face
Hugging Face enables image generation within Claude by connecting the AI to Hugging Face Spaces via the Model Context Protocol (MCP) server.
Qwen-Image-Edit Release Notes
Qwen-Image-Edit is a 20B parameter image editing model that enables precise semantic and appearance editing, including bilingual text modification, by leveraging Qwen2.5-VL and a VAE Encoder.
Hugging Face MCP for Research: Connecting AI to Research Tools
Hugging Face introduces the Research Tracker MCP, enabling AI agents to automate research discovery by integrating arXiv, GitHub, and Hugging Face through the Model Context Protocol.
Hugging Face kernel-builder: A Guide to Building and Scaling Production-Ready CUDA Kernels
Hugging Face introduces the kernel-builder library to simplify the development, multi-architecture compilation, and distribution of production-ready CUDA kernels via the Hugging Face Hub.
DoorDash AI Implementation Strategy
DoorDash is leveraging AI to democratize technical creation for non-engineers and personalize employee development and performance management.
Kimina-Prover-RL Release
Hugging Face has released Kimina-Prover-RL, an open-source RL training pipeline and two SOTA models (0.6B and 1.7B) for formal theorem proving in Lean 4.
Arm and ExecuTorch 0.7: Expanding Generative AI to Billions of Devices
Arm and the ExecuTorch 0.7 beta enable automatic AI acceleration via KleidiAI, leveraging the SDOT instruction to bring LLMs like Llama 3.2 to billions of existing Arm-based devices.
Arm Neural Super Sampling (NSS) Release
Arm has released Neural Super Sampling (NSS), an AI-powered upscaling solution designed to reduce GPU workloads and enable high-resolution rendering on mobile devices.
FilBench: Evaluating LLM Capabilities in Philippine Languages
Hugging Face has introduced FilBench, a comprehensive evaluation suite designed to assess the fluency, linguistic abilities, and cultural knowledge of LLMs in Tagalog, Filipino, and Cebuano.
TextQuests: Evaluating LLM Agentic Reasoning in Text-Based Video Games
Hugging Face introduces TextQuests, a benchmark using 25 classic Infocom interactive fiction games to evaluate the long-context reasoning and exploratory capabilities of LLM agents.
Basis Scales Accounting Automation with OpenAI GPT-5 and o3
Basis uses a multi-agent architecture powered by OpenAI's GPT-5, GPT-4.1, and o3 models to automate complex accounting workflows, achieving up to 30% time savings for accounting firms.
OpenAI Proposal for Harmonized AI Regulation in California
OpenAI has urged Governor Gavin Newsom to align California's AI regulations with national and global standards to prevent a patchwork of state rules that could hinder innovation and US competitiveness.
Hugging Face AI Sheets Release
Hugging Face has released AI Sheets, an open-source no-code tool for building, transforming, and enriching datasets using open AI models.
Accelerate ND-Parallel: Efficient Multi-GPU Training Guide
Hugging Face has integrated ND-Parallelism into Accelerate and Axolotl, allowing users to combine Data, Fully Sharded Data, Tensor, and Context parallelism strategies to optimize multi-GPU training for massive models.
OpenAI GPT-5 Release Notes
OpenAI has released GPT-5, a unified model that integrates reasoning, agents, and advanced math capabilities to improve accuracy, speed, and problem-solving for business operations.
GPT-5 for Developers Release Notes
OpenAI has released GPT-5 in three sizes (gpt-5, gpt-5-mini, and gpt-5-nano), delivering state-of-the-art performance in coding, agentic tool-calling, and long-context retrieval.
GPT-5 Coding and Design Capabilities
OpenAI has announced GPT-5 with a focus on enhanced capabilities for coding and design, though specific technical details were not provided in the announcement.
OpenAI GPT-5 Creative Writing Capabilities
OpenAI has announced new creative writing capabilities for GPT-5, expanding the model's ability to generate high-quality narrative and artistic text.
Medical Research with GPT-5
OpenAI has announced the application of GPT-5 to medical research, expanding the model's utility in specialized scientific domains.
Vision Language Model Alignment in TRL
Hugging Face has expanded the TRL library to support advanced alignment methods for Vision Language Models, including MPO, GRPO, and GSPO, alongside native SFT support and vLLM integration.
GPT-5 Integration in Cursor
OpenAI has announced the integration of GPT-5 into the Cursor code editor, enhancing AI-powered software development capabilities.
OpenAI GPT-5 First Look
OpenAI has provided a first look at GPT-5, marking a new generation of their large language model series.
How Amgen uses GPT-5
Amgen leverages GPT-5 via the OpenAI API to accelerate biotechnology research and drug discovery processes.
OpenAI GPT-5 System Card
OpenAI has released GPT-5, a unified system featuring a real-time router that dynamically switches between fast, high-throughput models and deeper reasoning models based on query complexity.
GPT-5 Safe Completions: Transitioning from Refusal-Based to Output-Centric Safety Training
OpenAI has introduced safe-completions for GPT-5, a safety-training approach that maximizes helpfulness while penalizing unsafe outputs, reducing the binary comply-or-refuse trade-off for dual-use prompts.
OpenAI GPT-5 Release Notes
OpenAI has released GPT-5, a unified AI system featuring a real-time router that switches between a fast model and a deeper reasoning model to provide expert-level intelligence across coding, math, and health.
OpenAI and GSA Partnership: ChatGPT Enterprise for U.S. Federal Workforce
OpenAI is providing ChatGPT Enterprise to the entire U.S. federal executive branch workforce for a nominal fee of $1 per agency for one year to reduce administrative burden and improve public service delivery.
OpenAI Estimating Worst Case Frontier Risks of Open Weight LLMs
OpenAI evaluated the risks of releasing gpt-oss by using malicious fine-tuning (MFT) to attempt to elicit maximum capabilities in biology and cybersecurity, finding that the model did not substantially advance the frontier of risk compared to existing open-weight models.
OpenAI gpt-oss-120b and gpt-oss-20b Release
OpenAI has released gpt-oss-120b and gpt-oss-20b, two open-weight reasoning models under the Apache 2.0 license designed for agentic workflows and customizable reasoning effort.